After spending some time with the new `flux kontext dev` model, I realized its most powerful capabilities aren't immediately obvious. Many people might miss its true potential by just scratching the surface.
I went deep and curated a collection of what I think are its most interesting use cases \u2013 things like targeted text removal, subtle photo restoration, and creative style transfers.
I felt that simply writing about them wasn't enough. The best way to understand the value is to see it and try it for yourself.
That's why I built FluxKontextLab (https://fluxkontextlab.com).
On the site, I've presented these curated examples with before-and-after comparisons. More importantly, there's an interactive playground right there, so you can immediately test these ideas or your own prompts on your own images.
My goal is to share what this model is capable of beyond the basics.
It's still an early project. I'd love for you to take a look and share your thoughts or any cool results you generate."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: I built a playground to showcase what Flux Kontext is good at"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://fluxkontextlab.com"}},"_tags":["story","author_Zephyrion","story_44516288","show_hn"],"author":"Zephyrion","children":[44516836,44516956,44517350,44518215,44518748,44518860,44519026,44525253,44530734,44531020,44531276,44560895],"created_at":"2025-07-10T01:10:08Z","created_at_i":1752109808,"num_comments":20,"objectID":"44516288","points":72,"story_id":44516288,"story_text":"Hi HN,
After spending some time with the new `flux kontext dev` model, I realized its most powerful capabilities aren't immediately obvious. Many people might miss its true potential by just scratching the surface.
I went deep and curated a collection of what I think are its most interesting use cases \u2013 things like targeted text removal, subtle photo restoration, and creative style transfers.
I felt that simply writing about them wasn't enough. The best way to understand the value is to see it and try it for yourself.
That's why I built FluxKontextLab (https://fluxkontextlab.com).
On the site, I've presented these curated examples with before-and-after comparisons. More importantly, there's an interactive playground right there, so you can immediately test these ideas or your own prompts on your own images.
My goal is to share what this model is capable of beyond the basics.
It's still an early project. I'd love for you to take a look and share your thoughts or any cool results you generate.","title":"Show HN: I built a playground to showcase what Flux Kontext is good at","updated_at":"2025-08-14T22:13:44Z","url":"https://fluxkontextlab.com"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mc-serious"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"We built the Kontext CLI because AI coding agents need access to GitHub, Stripe, databases, and dozens of other services \u2014 and right now most teams handle this by copy-pasting long-lived API keys into .env files, or the actual chat interface, whilst hoping for the best.
The problem isn't just secret sprawl. It's that there's no lineage of access. You don't know which developer launched which agent, what it accessed, or whether it should have been allowed to. The moment you hand raw credentials to a process, you've lost the ability to enforce policy, audit access, or rotate without pain. The credential is the authorization, and that's fundamentally broken when autonomous agents are making hundreds of API calls per session.
Kontext takes a different approach. You declare what credentials a project needs in a .env.kontext file:
GITHUB_TOKEN={{kontext:github}}\n STRIPE_KEY={{kontext:stripe}}\n LINEAR_TOKEN={{kontext:linear}}\nThe closest analogy is a Security Token Service (STS): you authenticate once, and the backend mints short-lived, scoped credentials on-the-fly \u2014 except unlike a classical STS, we hold the upstream secrets, so nothing long-lived ever reaches the agent. The backend holds your OAuth refresh tokens and API keys; the CLI never sees them. It gets back short-lived access tokens scoped to the session.
What the CLI captures for every tool call: what the agent tried to do, what happened, whether it was allowed, and who did it \u2014 attributed to a user, session, and org.
Install with one command: `brew install kontext-dev/tap/kontext`
The CLI is written in Go (~5ms hook overhead per tool call), uses ConnectRPC for backend communication, and stores auth in the system keyring. Works with Claude Code today, Codex support coming soon.
We're working on server-side policy enforcement next \u2014 the infrastructure for allow/deny decisions on every tool call is already wired, we just need to close the loop so tool calls can also be rejected.
We'd love feedback on the approach. Especially curious: how are teams handling credential management for AI agents today? Are you just pasting env vars into the agent chat, or have you found something better?
GitHub: https://github.com/kontext-dev/kontext-cli\nSite: https://kontext.security"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Kontext CLI \u2013 Credential broker for AI coding agents in Go"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://github.com/kontext-dev/kontext-cli"}},"_tags":["story","author_mc-serious","story_47765374","show_hn"],"author":"mc-serious","children":[47767439,47767447,47767609,47767892,47767955,47768069,47768168,47768933,47769350,47769489,47769510,47769988,47770187,47770396],"created_at":"2026-04-14T13:26:53Z","created_at_i":1776173213,"num_comments":17,"objectID":"47765374","points":70,"story_id":47765374,"story_text":"We built the Kontext CLI because AI coding agents need access to GitHub, Stripe, databases, and dozens of other services \u2014 and right now most teams handle this by copy-pasting long-lived API keys into .env files, or the actual chat interface, whilst hoping for the best.
The problem isn't just secret sprawl. It's that there's no lineage of access. You don't know which developer launched which agent, what it accessed, or whether it should have been allowed to. The moment you hand raw credentials to a process, you've lost the ability to enforce policy, audit access, or rotate without pain. The credential is the authorization, and that's fundamentally broken when autonomous agents are making hundreds of API calls per session.
Kontext takes a different approach. You declare what credentials a project needs in a .env.kontext file:
GITHUB_TOKEN={{kontext:github}}\n STRIPE_KEY={{kontext:stripe}}\n LINEAR_TOKEN={{kontext:linear}}\nThe closest analogy is a Security Token Service (STS): you authenticate once, and the backend mints short-lived, scoped credentials on-the-fly \u2014 except unlike a classical STS, we hold the upstream secrets, so nothing long-lived ever reaches the agent. The backend holds your OAuth refresh tokens and API keys; the CLI never sees them. It gets back short-lived access tokens scoped to the session.
What the CLI captures for every tool call: what the agent tried to do, what happened, whether it was allowed, and who did it \u2014 attributed to a user, session, and org.
Install with one command: `brew install kontext-dev/tap/kontext`
The CLI is written in Go (~5ms hook overhead per tool call), uses ConnectRPC for backend communication, and stores auth in the system keyring. Works with Claude Code today, Codex support coming soon.
We're working on server-side policy enforcement next \u2014 the infrastructure for allow/deny decisions on every tool call is already wired, we just need to close the loop so tool calls can also be rejected.
We'd love feedback on the approach. Especially curious: how are teams handling credential management for AI agents today? Are you just pasting env vars into the agent chat, or have you found something better?
GitHub: https://github.com/kontext-dev/kontext-cli\nSite: https://kontext.security","title":"Show HN: Kontext CLI \u2013 Credential broker for AI coding agents in Go","updated_at":"2026-05-07T01:55:13Z","url":"https://github.com/kontext-dev/kontext-cli"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"cyberplaid"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hi HN,
I created a photo restoration tool using the Flux Kontext model, and the results have been surprisingly good.
It can restore old photos, fix scratches and damage, and even colorize black-and-white images \u2014 all in just a few seconds.
It's currently completely free to use.
Would love to hear your feedback!"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: I built an old photo restoration tool using the Flux Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://restoreoldphotos.io"}},"_tags":["story","author_cyberplaid","story_44188672","show_hn"],"author":"cyberplaid","children":[44188994,44190504,44190756,44191540,44192011,44192102],"created_at":"2025-06-05T05:51:39Z","created_at_i":1749102699,"num_comments":8,"objectID":"44188672","points":16,"story_id":44188672,"story_text":"Hi HN,
I created a photo restoration tool using the Flux Kontext model, and the results have been surprisingly good.
It can restore old photos, fix scratches and damage, and even colorize black-and-white images \u2014 all in just a few seconds.
It's currently completely free to use.
Would love to hear your feedback!","title":"Show HN: I built an old photo restoration tool using the Flux Kontext","updated_at":"2025-06-07T13:49:01Z","url":"https://restoreoldphotos.io"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"anuragmerndev"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Kontext \u2013 Move an AI chat's full context to another AI in one click"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://github.com/anuragmerndev/kontext-ai"}},"_tags":["story","author_anuragmerndev","story_48778955","show_hn"],"author":"anuragmerndev","children":[48778962,48780896,48785329],"created_at":"2026-07-03T19:27:04Z","created_at_i":1783106824,"num_comments":2,"objectID":"48778955","points":8,"story_id":48778955,"title":"Show HN: Kontext \u2013 Move an AI chat's full context to another AI in one click","updated_at":"2026-07-05T17:43:16Z","url":"https://github.com/anuragmerndev/kontext-ai"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"michiosw"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Every AI agent that does something useful - opening a PR, posting in Slack, updating a ticket - needs to call an API on behalf of a user. That means OAuth. Authorization flows, token storage, refresh logic, per-user credential isolation.
Today, most teams solve this with a long-lived API key in an .env file, shared across every user and every session. As everyone in an organization becomes a software engineer - whether they know it or not - you can't expect each of them to roll their own OAuth. Secure credential handling needs to be baked into the infrastructure.
The core issue: agents are a new kind of principal. They're not users. They're not service accounts. They act with delegated authority on someone else's behalf. Existing identity systems have no model for this.
Kontext does.
Kontext validates the user and the agent, and returns a credential that lets agents act on behalf of their user - scoped, short-lived, and revocable. No secrets in your codebase, and action is traced back to the human who approved it.
We are thrilled to be launching our SDK, let us know what you think!"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Kontext.dev \u2013 Runtime Credentials for Agents"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://kontext.dev/blog/announcing-kontext"}},"_tags":["story","author_michiosw","story_47401302","show_hn"],"author":"michiosw","children":[47401322,47405774,47406786,47412100,47453332],"created_at":"2026-03-16T16:39:54Z","created_at_i":1773679194,"num_comments":1,"objectID":"47401302","points":6,"story_id":47401302,"story_text":"Every AI agent that does something useful - opening a PR, posting in Slack, updating a ticket - needs to call an API on behalf of a user. That means OAuth. Authorization flows, token storage, refresh logic, per-user credential isolation.
Today, most teams solve this with a long-lived API key in an .env file, shared across every user and every session. As everyone in an organization becomes a software engineer - whether they know it or not - you can't expect each of them to roll their own OAuth. Secure credential handling needs to be baked into the infrastructure.
The core issue: agents are a new kind of principal. They're not users. They're not service accounts. They act with delegated authority on someone else's behalf. Existing identity systems have no model for this.
Kontext does.
Kontext validates the user and the agent, and returns a credential that lets agents act on behalf of their user - scoped, short-lived, and revocable. No secrets in your codebase, and action is traced back to the human who approved it.
We are thrilled to be launching our SDK, let us know what you think!","title":"Show HN: Kontext.dev \u2013 Runtime Credentials for Agents","updated_at":"2026-03-27T03:01:38Z","url":"https://kontext.dev/blog/announcing-kontext"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"sanaka87"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"27 GPU hours, BAGEL with self-supervised post-training beats FLUX-Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.alphaxiv.org/abs/2509.07295v1"}},"_tags":["story","author_sanaka87","story_45228773"],"author":"sanaka87","created_at":"2025-09-13T02:14:43Z","created_at_i":1757729683,"num_comments":0,"objectID":"45228773","points":6,"story_id":45228773,"title":"27 GPU hours, BAGEL with self-supervised post-training beats FLUX-Kontext","updated_at":"2026-03-05T22:40:13Z","url":"https://www.alphaxiv.org/abs/2509.07295v1"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"dberenstein1957"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: FLUX.1 Kontext[DEV]-juiced SOTA Image Editing but 5x Faster"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://www.pruna.ai/blog/flux-kontext-juiced-state-of-the-art-image-editing-x5-faster"}},"_tags":["story","author_dberenstein1957","story_44388653","show_hn"],"author":"dberenstein1957","created_at":"2025-06-26T15:59:06Z","created_at_i":1750953546,"num_comments":0,"objectID":"44388653","points":4,"story_id":44388653,"title":"Show HN: FLUX.1 Kontext[DEV]-juiced SOTA Image Editing but 5x Faster","updated_at":"2025-06-27T13:22:47Z","url":"https://www.pruna.ai/blog/flux-kontext-juiced-state-of-the-art-image-editing-x5-faster"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"stephen37"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"FLUX.1 Kontext \u2013 Edit Images with Prompt"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://bfl.ai/announcements/flux-1-kontext"}},"_tags":["story","author_stephen37","story_44128458"],"author":"stephen37","created_at":"2025-05-29T17:53:57Z","created_at_i":1748541237,"num_comments":0,"objectID":"44128458","points":4,"story_id":44128458,"title":"FLUX.1 Kontext \u2013 Edit Images with Prompt","updated_at":"2025-05-29T18:29:56Z","url":"https://bfl.ai/announcements/flux-1-kontext"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"noshbrinken"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: kontext \u2013 Context in JavaScript functions without `this`"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://github.com/maxhallinan/kontext"}},"_tags":["story","author_noshbrinken","story_15399103","show_hn"],"author":"noshbrinken","created_at":"2017-10-04T08:53:17Z","created_at_i":1507107197,"num_comments":0,"objectID":"15399103","points":4,"story_id":15399103,"title":"Show HN: kontext \u2013 Context in JavaScript functions without `this`","updated_at":"2024-09-20T01:31:16Z","url":"https://github.com/maxhallinan/kontext"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"amrrs"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"FLUX.1 Kontext [Dev] Inference and Training"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://blog.fal.ai/announcing-flux-1-kontext-dev-inference-training/"}},"_tags":["story","author_amrrs","story_44389980"],"author":"amrrs","children":[44389990],"created_at":"2025-06-26T18:28:43Z","created_at_i":1750962523,"num_comments":0,"objectID":"44389980","points":3,"story_id":44389980,"title":"FLUX.1 Kontext [Dev] Inference and Training","updated_at":"2025-06-26T18:49:15Z","url":"https://blog.fal.ai/announcing-flux-1-kontext-dev-inference-training/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"chengzeyi"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"FLUX.1 Kontext is a new image editing model from Black Forest Labs. It is the best in class model for editing images using text prompts, and the latest addition to the FLUX.1 family.
In our tests we\u2019ve found Kontext to give accurate and brilliant results. It\u2019s better and cheaper than OpenAI\u2019s 4o/gpt-image-1 model (and there\u2019s no yellow tint).
There are three models, two are available now, and a third open-weight version is coming soon:
FLUX.1 Kontext [pro]: State-of-the-art performance for image editing. High-quality outputs, great prompt following, and consistent results.\nFLUX.1 Kontext [max]: A premium model that brings maximum performance, improved prompt adherence, and high-quality typography generation without compromise on speed.\nComing soon: FLUX.1 Kontext [dev]: An open-weight, guidance-distilled version of Kontext."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Flux Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.fluxkontext.com/"}},"_tags":["story","author_chengzeyi","story_44131871","show_hn"],"author":"chengzeyi","created_at":"2025-05-30T01:13:53Z","created_at_i":1748567633,"num_comments":0,"objectID":"44131871","points":3,"story_id":44131871,"story_text":"FLUX.1 Kontext is a new image editing model from Black Forest Labs. It is the best in class model for editing images using text prompts, and the latest addition to the FLUX.1 family.
In our tests we\u2019ve found Kontext to give accurate and brilliant results. It\u2019s better and cheaper than OpenAI\u2019s 4o/gpt-image-1 model (and there\u2019s no yellow tint).
There are three models, two are available now, and a third open-weight version is coming soon:
FLUX.1 Kontext [pro]: State-of-the-art performance for image editing. High-quality outputs, great prompt following, and consistent results.\nFLUX.1 Kontext [max]: A premium model that brings maximum performance, improved prompt adherence, and high-quality typography generation without compromise on speed.\nComing soon: FLUX.1 Kontext [dev]: An open-weight, guidance-distilled version of Kontext.","title":"Show HN: Flux Kontext","updated_at":"2025-06-04T01:18:31Z","url":"https://www.fluxkontext.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"anjneymidha"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"FLUX.1 Kontext:In-Context Image Generation and Editing in Latent Space [pdf]"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://cdn.sanity.io/files/gsvmb6gz/production/880b072208997108f87e5d2729d8a8be481310b5.pdf"}},"_tags":["story","author_anjneymidha","story_44128678"],"author":"anjneymidha","created_at":"2025-05-29T18:15:18Z","created_at_i":1748542518,"num_comments":0,"objectID":"44128678","points":3,"story_id":44128678,"title":"FLUX.1 Kontext:In-Context Image Generation and Editing in Latent Space [pdf]","updated_at":"2025-05-29T18:20:12Z","url":"https://cdn.sanity.io/files/gsvmb6gz/production/880b072208997108f87e5d2729d8a8be481310b5.pdf"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"p_sekhar"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Use FLUX.1 Kontext to edit images with words"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://replicate.com/blog/flux-kontext"}},"_tags":["story","author_p_sekhar","story_44128258"],"author":"p_sekhar","created_at":"2025-05-29T17:33:49Z","created_at_i":1748540029,"num_comments":0,"objectID":"44128258","points":3,"story_id":44128258,"title":"Use FLUX.1 Kontext to edit images with words","updated_at":"2025-05-30T09:14:14Z","url":"https://replicate.com/blog/flux-kontext"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"zzsshh"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Built this to show off the recently released Flux Kontext model which is great at making edits based on your description, while keeping everything else the same. Eg. "make him wear a suit", "change the background to the beach", etc. No mask, etc needed.
I've used Nextjs 15, and this page is static and fully client side. I'm still torn between the value of RSC, and the downside of it not being served from a CDN.
I use Supabase for Db and Auth. Happy with these general, but their auth needs some work. The official supabase auth docs still recommend logic in nextjs middleware which hits your DB for protected routes.
My app gives you 3 free edits per day with no sign up required. There's rudimentary vibe-coded logic to enforce this, please don't bankrupt me, i have a wife and kids.
There's potential to expand on what the app can do with flux kontext also supporting multiple images as a prompt, for things like combining people into one photo.
Open to feedback and ideas on the tech, product or pricing."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: I built a web app to edit photos with a prompt using Flux Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.theinfluencer.ai/ai-photo-editor"}},"_tags":["story","author_zzsshh","story_44309782","show_hn"],"author":"zzsshh","children":[44312094],"created_at":"2025-06-18T13:37:00Z","created_at_i":1750253820,"num_comments":2,"objectID":"44309782","points":2,"story_id":44309782,"story_text":"Built this to show off the recently released Flux Kontext model which is great at making edits based on your description, while keeping everything else the same. Eg. "make him wear a suit", "change the background to the beach", etc. No mask, etc needed.
I've used Nextjs 15, and this page is static and fully client side. I'm still torn between the value of RSC, and the downside of it not being served from a CDN.
I use Supabase for Db and Auth. Happy with these general, but their auth needs some work. The official supabase auth docs still recommend logic in nextjs middleware which hits your DB for protected routes.
My app gives you 3 free edits per day with no sign up required. There's rudimentary vibe-coded logic to enforce this, please don't bankrupt me, i have a wife and kids.
There's potential to expand on what the app can do with flux kontext also supporting multiple images as a prompt, for things like combining people into one photo.
Open to feedback and ideas on the tech, product or pricing.","title":"Show HN: I built a web app to edit photos with a prompt using Flux Kontext","updated_at":"2025-06-18T19:12:43Z","url":"https://www.theinfluencer.ai/ai-photo-editor"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"VidText_AI"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"How powerful is FLUX.1 Kontext?"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://fluxkontext.top"}},"_tags":["story","author_VidText_AI","story_44167630"],"author":"VidText_AI","children":[44167631,44167770],"created_at":"2025-06-03T08:11:14Z","created_at_i":1748938274,"num_comments":2,"objectID":"44167630","points":2,"story_id":44167630,"title":"How powerful is FLUX.1 Kontext?","updated_at":"2025-06-03T12:41:44Z","url":"https://fluxkontext.top"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"kfarr"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"From low-poly to animated urban design renders using Flux Kontext Pro & Veo3"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.3dstreet.com/blog/2025/07/15/creating-photorealistic-pictures-and-animations-with-3dstreet-ai/"}},"_tags":["story","author_kfarr","story_44659189"],"author":"kfarr","children":[44659190],"created_at":"2025-07-23T13:45:21Z","created_at_i":1753278321,"num_comments":1,"objectID":"44659189","points":2,"story_id":44659189,"title":"From low-poly to animated urban design renders using Flux Kontext Pro & Veo3","updated_at":"2025-07-23T19:30:16Z","url":"https://www.3dstreet.com/blog/2025/07/15/creating-photorealistic-pictures-and-animations-with-3dstreet-ai/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"XMasterrrr"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"I Built My Wife a Simple Web App for Image Editing Using Flux Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/TheAhmadOsman/4o-ghibli-at-home"}},"_tags":["story","author_XMasterrrr","story_44448044"],"author":"XMasterrrr","children":[44449003],"created_at":"2025-07-02T19:49:30Z","created_at_i":1751485770,"num_comments":1,"objectID":"44448044","points":2,"story_id":44448044,"title":"I Built My Wife a Simple Web App for Image Editing Using Flux Kontext","updated_at":"2025-07-02T21:23:38Z","url":"https://github.com/TheAhmadOsman/4o-ghibli-at-home"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"myth_drannon"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"FLUX.1 Kontext and the BFL Playground"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://bfl.ai/announcements/flux-1-kontext"}},"_tags":["story","author_myth_drannon","story_44398617"],"author":"myth_drannon","children":[44398627,44398628],"created_at":"2025-06-27T17:27:12Z","created_at_i":1751045232,"num_comments":1,"objectID":"44398617","points":2,"story_id":44398617,"title":"FLUX.1 Kontext and the BFL Playground","updated_at":"2025-06-29T02:52:23Z","url":"https://bfl.ai/announcements/flux-1-kontext"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"chengzeyi"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"FLUX Kontext Dev Ultra Fast Live"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://huggingface.co/spaces/wavespeed/FLUX-Kontext-Dev-Ultra-Fast"}},"_tags":["story","author_chengzeyi","story_44388370"],"author":"chengzeyi","children":[44388371,44388399],"created_at":"2025-06-26T15:29:33Z","created_at_i":1750951773,"num_comments":1,"objectID":"44388370","points":2,"story_id":44388370,"title":"FLUX Kontext Dev Ultra Fast Live","updated_at":"2025-06-26T15:39:59Z","url":"https://huggingface.co/spaces/wavespeed/FLUX-Kontext-Dev-Ultra-Fast"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"AllenRen"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"FLUX Kontext: AI-Powered Image Editing Made Simple\nHey HN,\nI've been working on https://kontextflux.com \u2013 an AI-powered image editing platform using Black Forest Labs' FLUX.1 Kontext models, designed for easy, professional-grade visual edits without requiring expert skills.
What it does\uff1aUpload an image, describe your desired changes in plain text, and it edits only the specified parts while preserving the rest. Think "change the character's expression" or "add a new background style" \u2013 it understands context and maintains consistency.
Tech stack\uff1a\n1.Next.js 19, TypeScript\n2.Supabase for database\n3.FAL.AI for model hosting\n4.Cloudflare R2 for storageStandard SaaS setup optimized for performance.
Why I built this\uff1f\nMost AI image tools either regenerate entire images or produce inconsistent edits. FLUX Kontext excels at precise, context-aware edits, preserving the original image\u2019s style, lighting, and composition while supporting multi-step edits and style consistency.
Current state\uff1a\nSupports common image formats (up to 10MB)\nProcessing time: ~10-45 seconds depending on complexity
Key features: \nText-to-image fusion and multi-image blending\nPrecise local editing via natural language\nCharacter style consistency across scenes\nNear real-time feedback for iterative edits
Challenges\uff1a\nModel hosting costs are significant, so optimization is key\nFAL.AI API can lag during peak times\nBalancing speed and quality for the best user experience
This platform is ideal for digital artists, graphic designers, social media operators, film/game concept designers, and e-commerce teams who need fast, consistent, and professional-grade image edits without complex workflows. The tech, built by the Stable Diffusion team, delivers impressive results. You can try the demo without signing up!"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"FLUX.1 Kontext \u2014Context-aware image generation and editing tools"}},"_tags":["story","author_AllenRen","story_44169890","ask_hn"],"author":"AllenRen","children":[44188788,44188790],"created_at":"2025-06-03T13:29:56Z","created_at_i":1748957396,"num_comments":1,"objectID":"44169890","points":2,"story_id":44169890,"story_text":"FLUX Kontext: AI-Powered Image Editing Made Simple\nHey HN,\nI've been working on https://kontextflux.com \u2013 an AI-powered image editing platform using Black Forest Labs' FLUX.1 Kontext models, designed for easy, professional-grade visual edits without requiring expert skills.
What it does\uff1aUpload an image, describe your desired changes in plain text, and it edits only the specified parts while preserving the rest. Think "change the character's expression" or "add a new background style" \u2013 it understands context and maintains consistency.
Tech stack\uff1a\n1.Next.js 19, TypeScript\n2.Supabase for database\n3.FAL.AI for model hosting\n4.Cloudflare R2 for storageStandard SaaS setup optimized for performance.
Why I built this\uff1f\nMost AI image tools either regenerate entire images or produce inconsistent edits. FLUX Kontext excels at precise, context-aware edits, preserving the original image\u2019s style, lighting, and composition while supporting multi-step edits and style consistency.
Current state\uff1a\nSupports common image formats (up to 10MB)\nProcessing time: ~10-45 seconds depending on complexity
Key features: \nText-to-image fusion and multi-image blending\nPrecise local editing via natural language\nCharacter style consistency across scenes\nNear real-time feedback for iterative edits
Challenges\uff1a\nModel hosting costs are significant, so optimization is key\nFAL.AI API can lag during peak times\nBalancing speed and quality for the best user experience
This platform is ideal for digital artists, graphic designers, social media operators, film/game concept designers, and e-commerce teams who need fast, consistent, and professional-grade image edits without complex workflows. The tech, built by the Stable Diffusion team, delivers impressive results. You can try the demo without signing up!","title":"FLUX.1 Kontext \u2014Context-aware image generation and editing tools","updated_at":"2025-06-05T06:13:21Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"wushi"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hey HN,
I've been working on https://flux1kontext.io \u2013 an AI image editor that uses Black Forest Labs' FLUX.1 Kontext models for context-aware image editing.
*What it does:* Upload an image, describe what you want to change, and it modifies just that part while keeping everything else intact. Think "add sunglasses to this person" or "make the cat open its eyes" \u2013 it understands context instead of regenerating the whole image.
*Tech stack:* Next.js 19, TypeScript, Supabase, FAL.AI for model hosting, Stripe for payments, Cloudflare R2 for storage. Pretty standard SaaS setup.
*Why I built this:* Most AI image tools either generate from scratch or do crude inpainting. FLUX.1 Kontext actually understands what you're asking for and preserves the original image's composition, lighting, and style while making targeted edits.
*Current state:* \n- Two model tiers (Pro/Max) with different quality/speed tradeoffs\n- Credit-based pricing (8 credits for Pro, 16 for Max)\n- Works with common image formats, handles up to 10MB files\n- About 15-60 seconds processing time depending on complexity
*Challenges:* \n- Model costs are high (~$0.10-0.20 per generation), so had to implement careful credit management\n- FAL.AI API can be inconsistent during peak times\n- Balancing quality vs speed is tricky \u2013 users want both
The tech is genuinely impressive when it works. The model was trained by the same team that created Stable Diffusion, so the quality is solid. Still working on reliability and cost optimization.
Would love feedback from anyone who's worked with similar AI APIs or has thoughts on the UX. The demo works without signup if you want to try it.
Built this mostly as a learning project but it's getting decent usage (~2.5M images generated so far). Happy to answer questions about the implementation or business side."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Built FLUX.1 Kontext image editor \u2013 context-aware AI editing"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://flux1kontext.io"}},"_tags":["story","author_wushi","story_44149393","show_hn"],"author":"wushi","children":[44150022],"created_at":"2025-06-01T08:15:27Z","created_at_i":1748765727,"num_comments":1,"objectID":"44149393","points":2,"story_id":44149393,"story_text":"Hey HN,
I've been working on https://flux1kontext.io \u2013 an AI image editor that uses Black Forest Labs' FLUX.1 Kontext models for context-aware image editing.
*What it does:* Upload an image, describe what you want to change, and it modifies just that part while keeping everything else intact. Think "add sunglasses to this person" or "make the cat open its eyes" \u2013 it understands context instead of regenerating the whole image.
*Tech stack:* Next.js 19, TypeScript, Supabase, FAL.AI for model hosting, Stripe for payments, Cloudflare R2 for storage. Pretty standard SaaS setup.
*Why I built this:* Most AI image tools either generate from scratch or do crude inpainting. FLUX.1 Kontext actually understands what you're asking for and preserves the original image's composition, lighting, and style while making targeted edits.
*Current state:* \n- Two model tiers (Pro/Max) with different quality/speed tradeoffs\n- Credit-based pricing (8 credits for Pro, 16 for Max)\n- Works with common image formats, handles up to 10MB files\n- About 15-60 seconds processing time depending on complexity
*Challenges:* \n- Model costs are high (~$0.10-0.20 per generation), so had to implement careful credit management\n- FAL.AI API can be inconsistent during peak times\n- Balancing quality vs speed is tricky \u2013 users want both
The tech is genuinely impressive when it works. The model was trained by the same team that created Stable Diffusion, so the quality is solid. Still working on reliability and cost optimization.
Would love feedback from anyone who's worked with similar AI APIs or has thoughts on the UX. The demo works without signup if you want to try it.
Built this mostly as a learning project but it's getting decent usage (~2.5M images generated so far). Happy to answer questions about the implementation or business side.","title":"Show HN: Built FLUX.1 Kontext image editor \u2013 context-aware AI editing","updated_at":"2025-06-01T11:07:51Z","url":"https://flux1kontext.io"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"tdrnd"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Kontext: A context-shift transition"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"http://lab.hakim.se/kontext/"}},"_tags":["story","author_tdrnd","story_5286237"],"author":"tdrnd","created_at":"2013-02-26T15:47:23Z","created_at_i":1361893643,"num_comments":0,"objectID":"5286237","points":2,"story_id":5286237,"story_text":"","title":"Kontext: A context-shift transition","updated_at":"2024-09-19T19:21:15Z","url":"http://lab.hakim.se/kontext/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"waltercool"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Flux.1 Kontext [Dev] Day-0 Support in ComfyUI"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://blog.comfy.org/p/flux1-kontext-dev-day-0-support"}},"_tags":["story","author_waltercool","story_44416828"],"author":"waltercool","created_at":"2025-06-29T21:56:09Z","created_at_i":1751234169,"num_comments":0,"objectID":"44416828","points":2,"story_id":44416828,"title":"Flux.1 Kontext [Dev] Day-0 Support in ComfyUI","updated_at":"2025-06-29T22:03:58Z","url":"https://blog.comfy.org/p/flux1-kontext-dev-day-0-support"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"lexandstuff"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Absurdly Good Doggo Consistency with FLUX.1 Kontext"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://notesbylex.com/absurdly-good-doggo-consistency-with-flux1-kontext"}},"_tags":["story","author_lexandstuff","story_44154415"],"author":"lexandstuff","children":[44154749],"created_at":"2025-06-01T23:04:59Z","created_at_i":1748819099,"num_comments":0,"objectID":"44154415","points":2,"story_id":44154415,"title":"Absurdly Good Doggo Consistency with FLUX.1 Kontext","updated_at":"2025-06-02T00:32:23Z","url":"https://notesbylex.com/absurdly-good-doggo-consistency-with-flux1-kontext"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"taherchhabra"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Flux Kontext Image editing tests"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.flickspeed.ai/canvas/public/6871319e239a5c68830ee64f"}},"_tags":["story","author_taherchhabra","story_44536358"],"author":"taherchhabra","children":[44536359],"created_at":"2025-07-11T20:09:59Z","created_at_i":1752264599,"num_comments":1,"objectID":"44536358","points":1,"story_id":44536358,"title":"Flux Kontext Image editing tests","updated_at":"2025-07-11T20:11:42Z","url":"https://www.flickspeed.ai/canvas/public/6871319e239a5c68830ee64f"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"lumen2088"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Flux.1 Kontext Free Online: AI-Powered Image Editing"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://iflux.ai"}},"_tags":["story","author_lumen2088","story_44422823"],"author":"lumen2088","children":[44422824],"created_at":"2025-06-30T13:07:12Z","created_at_i":1751288832,"num_comments":1,"objectID":"44422823","points":1,"story_id":44422823,"title":"Flux.1 Kontext Free Online: AI-Powered Image Editing","updated_at":"2025-06-30T13:08:58Z","url":"https://iflux.ai"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"matthewshere"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"AI Image Editor Powered by Flux Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://fluxkontextai.pro/"}},"_tags":["story","author_matthewshere","story_44245401"],"author":"matthewshere","children":[44245402],"created_at":"2025-06-11T08:27:33Z","created_at_i":1749630453,"num_comments":1,"objectID":"44245401","points":1,"story_id":44245401,"title":"AI Image Editor Powered by Flux Kontext","updated_at":"2025-06-11T14:42:31Z","url":"https://fluxkontextai.pro/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"zhaomeng"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Flux Kontext AI"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://flux1kontext.ai/"}},"_tags":["story","author_zhaomeng","story_45780530"],"author":"zhaomeng","children":[45780531],"created_at":"2025-11-01T10:11:55Z","created_at_i":1761991915,"num_comments":0,"objectID":"45780530","points":1,"story_id":45780530,"title":"Flux Kontext AI","updated_at":"2026-03-05T22:59:47Z","url":"https://flux1kontext.ai/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"mehelpme"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"We\u2019ve built Flux Kontext \u2014 an AI image editing platform that allows LoRA and ControlNet editing, character consistency, and restoration tools.\nYou can test it here: https://flux1kontextai.com\nFeedback is welcome!"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Flux Kontext \u2013 AI-Powered Image Restoration Platform"}},"_tags":["story","author_mehelpme","story_45755569","ask_hn"],"author":"mehelpme","created_at":"2025-10-30T01:56:03Z","created_at_i":1761789363,"num_comments":0,"objectID":"45755569","points":1,"story_id":45755569,"story_text":"We\u2019ve built Flux Kontext \u2014 an AI image editing platform that allows LoRA and ControlNet editing, character consistency, and restoration tools.\nYou can test it here: https://flux1kontextai.com\nFeedback is welcome!","title":"Flux Kontext \u2013 AI-Powered Image Restoration Platform","updated_at":"2026-03-05T22:58:18Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"incendies"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hey HN!\nI just launched PHOAIART, an AI-powered photo editor where users can edit their images simply by chatting\u2014no buttons, no sliders, just natural language. You can say things like \u201cchange my hair to red\u201d or \u201cremove the background,\u201d and the AI gets it done instantly.
I built it using Flux Kontext for natural language image transformation, along with Firebase (Firestore + Auth) and React Native for the core app logic. The goal was to make image editing feel as easy as texting a friend.
Would love your feedback, suggestions, or ideas for improvement!
iOS App: https://apps.apple.com/app/phoaiart-ai-photo-editor/id674789..."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Phoaiart \u2013 Chat-Based AI Photo Editor Built with Flux Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://apps.apple.com/us/app/phoaiart-ai-photo-editor/id6747897909"}},"_tags":["story","author_incendies","story_44587703","show_hn"],"author":"incendies","created_at":"2025-07-16T22:54:54Z","created_at_i":1752706494,"num_comments":0,"objectID":"44587703","points":1,"story_id":44587703,"story_text":"Hey HN!\nI just launched PHOAIART, an AI-powered photo editor where users can edit their images simply by chatting\u2014no buttons, no sliders, just natural language. You can say things like \u201cchange my hair to red\u201d or \u201cremove the background,\u201d and the AI gets it done instantly.
I built it using Flux Kontext for natural language image transformation, along with Firebase (Firestore + Auth) and React Native for the core app logic. The goal was to make image editing feel as easy as texting a friend.
Would love your feedback, suggestions, or ideas for improvement!
iOS App: https://apps.apple.com/app/phoaiart-ai-photo-editor/id674789...","title":"Show HN: Phoaiart \u2013 Chat-Based AI Photo Editor Built with Flux Kontext","updated_at":"2025-07-16T22:59:04Z","url":"https://apps.apple.com/us/app/phoaiart-ai-photo-editor/id6747897909"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"bixantil"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"AI Product Image Suite with Flux Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://snap-z.com/"}},"_tags":["story","author_bixantil","story_44525959"],"author":"bixantil","children":[44525960],"created_at":"2025-07-10T21:44:36Z","created_at_i":1752183876,"num_comments":0,"objectID":"44525959","points":1,"story_id":44525959,"title":"AI Product Image Suite with Flux Kontext","updated_at":"2025-07-10T21:49:09Z","url":"https://snap-z.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"gregzeng95"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Flux Kontext AI \u2013 AI Image Generator and Editor"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://fluxkontext.io/"}},"_tags":["story","author_gregzeng95","story_44288023","show_hn"],"author":"gregzeng95","created_at":"2025-06-16T10:02:35Z","created_at_i":1750068155,"num_comments":0,"objectID":"44288023","points":1,"story_id":44288023,"title":"Show HN: Flux Kontext AI \u2013 AI Image Generator and Editor","updated_at":"2025-06-16T10:05:48Z","url":"https://fluxkontext.io/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"xbaicai"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Kontext-Chat"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://chat-photo.com"}},"_tags":["story","author_xbaicai","story_44235013"],"author":"xbaicai","children":[44235014],"created_at":"2025-06-10T10:33:33Z","created_at_i":1749551613,"num_comments":0,"objectID":"44235013","points":1,"story_id":44235013,"title":"Kontext-Chat","updated_at":"2025-06-10T10:38:26Z","url":"https://chat-photo.com"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"akseli_ukkonen"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"I was using Flux.1 Kontext for image editing but kept running into small annoyances, so I built some improvements on top of it.
Here's what I added:
- You can draw simple sketches directly on the image to show exactly what you want to change. Works better than just describing it in text.
- It automatically translates your prompts to English since Flux.1 Kontext only understands English. I got tired of copy-pasting into Google Translate every time.
- It tries to improve your prompts automatically. Like if you write "change her hair" it'll suggest "change the woman's hair" instead, or if you want to edit text it'll add quotes around it.
I'm sharing the exact prompt I use for optimization in case anyone finds it useful:
"Please analyze and optimize the following image editing prompt. Follow these guidelines:
1. Consistency Enhancement: Only add consistency maintenance descriptions when the prompt mentions specific regions/features/objects to modify AND does not already mention keeping other parts unchanged. In such cases, add descriptions like "maintain consistency with the original image except for the requested modifications".
2. Simplification: If the modification involves complex steps, break down complex edits into smaller, manageable steps. Larger modifications are easier to manage when done sequentially.
3. Pronoun Optimization: Replace vague references like "she", "her" with accurate references like "the woman in the picture".
4. Text Quotation: If the user wants to edit text but hasn't enclosed the text in quotes, add quotes around the text they describe. For example: "replace x into y" should become "replace 'x' into 'y'".
5. Accurate Description: If the user uses broad descriptions, change them to more accurate descriptions while understanding the user's original intent.
6. Structure Improvement: Ensure the prompt is well-structured and clear in its instructions.
7. Language Translation: If the input text is not in English, translate it to English while preserving the original intent and applying all the above optimizations.
Input prompt to optimize: <text-start> "${prompt}" <text-end>
Return the final optimized prompt in English that incorporates all these improvements."
Feel free to use this prompt however you'd like \u2013 it's completely open for anyone to use.
It's at https://kontextart.com if you want to try it. New signups get some free credits to test it out."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Kontext Art \u2013 I added some practical features to Flux.1 Kontext"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://kontextart.com"}},"_tags":["story","author_akseli_ukkonen","story_44211700","show_hn"],"author":"akseli_ukkonen","created_at":"2025-06-07T18:50:55Z","created_at_i":1749322255,"num_comments":0,"objectID":"44211700","points":1,"story_id":44211700,"story_text":"I was using Flux.1 Kontext for image editing but kept running into small annoyances, so I built some improvements on top of it.
Here's what I added:
- You can draw simple sketches directly on the image to show exactly what you want to change. Works better than just describing it in text.
- It automatically translates your prompts to English since Flux.1 Kontext only understands English. I got tired of copy-pasting into Google Translate every time.
- It tries to improve your prompts automatically. Like if you write "change her hair" it'll suggest "change the woman's hair" instead, or if you want to edit text it'll add quotes around it.
I'm sharing the exact prompt I use for optimization in case anyone finds it useful:
"Please analyze and optimize the following image editing prompt. Follow these guidelines:
1. Consistency Enhancement: Only add consistency maintenance descriptions when the prompt mentions specific regions/features/objects to modify AND does not already mention keeping other parts unchanged. In such cases, add descriptions like "maintain consistency with the original image except for the requested modifications".
2. Simplification: If the modification involves complex steps, break down complex edits into smaller, manageable steps. Larger modifications are easier to manage when done sequentially.
3. Pronoun Optimization: Replace vague references like "she", "her" with accurate references like "the woman in the picture".
4. Text Quotation: If the user wants to edit text but hasn't enclosed the text in quotes, add quotes around the text they describe. For example: "replace x into y" should become "replace 'x' into 'y'".
5. Accurate Description: If the user uses broad descriptions, change them to more accurate descriptions while understanding the user's original intent.
6. Structure Improvement: Ensure the prompt is well-structured and clear in its instructions.
7. Language Translation: If the input text is not in English, translate it to English while preserving the original intent and applying all the above optimizations.
Input prompt to optimize: <text-start> "${prompt}" <text-end>
Return the final optimized prompt in English that incorporates all these improvements."
Feel free to use this prompt however you'd like \u2013 it's completely open for anyone to use.
It's at https://kontextart.com if you want to try it. New signups get some free credits to test it out.","title":"Show HN: Kontext Art \u2013 I added some practical features to Flux.1 Kontext","updated_at":"2025-06-07T18:56:16Z","url":"https://kontextart.com"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"useapi"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"LTX Studio API v1 Featuring LTX-Video and FLUX.1 Kontext"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://useapi.net/blog/250603"}},"_tags":["story","author_useapi","story_44197509"],"author":"useapi","created_at":"2025-06-06T03:08:29Z","created_at_i":1749179309,"num_comments":0,"objectID":"44197509","points":1,"story_id":44197509,"title":"LTX Studio API v1 Featuring LTX-Video and FLUX.1 Kontext","updated_at":"2025-06-06T03:11:24Z","url":"https://useapi.net/blog/250603"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Sawana"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hey HN,
I built Kontext AI, a web app that leverages the FLUX.1 Kontext model for in-context image generation and editing.
I built this with Next.js (App Router), TypeScript, Tailwind CSS, and integrated with Replicate and Clerk for authentication. The app is fully responsive and offers several pricing tiers including a free option to try it out.
Would love feedback on the UI/UX, generation quality, feature request, and overall concept!
Try it out (easy sign-up with google one tap sign-in): https://kontextai.run/"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: Kontext AI \u2013 In-Context Image Generation and Editing with FLUX.1"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://kontextai.run"}},"_tags":["story","author_Sawana","story_44169823","show_hn"],"author":"Sawana","created_at":"2025-06-03T13:23:23Z","created_at_i":1748957003,"num_comments":0,"objectID":"44169823","points":1,"story_id":44169823,"story_text":"Hey HN,
I built Kontext AI, a web app that leverages the FLUX.1 Kontext model for in-context image generation and editing.
I built this with Next.js (App Router), TypeScript, Tailwind CSS, and integrated with Replicate and Clerk for authentication. The app is fully responsive and offers several pricing tiers including a free option to try it out.
Would love feedback on the UI/UX, generation quality, feature request, and overall concept!
Try it out (easy sign-up with google one tap sign-in): https://kontextai.run/","title":"Show HN: Kontext AI \u2013 In-Context Image Generation and Editing with FLUX.1","updated_at":"2025-06-03T13:24:29Z","url":"https://kontextai.run"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"AllenRen"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Flux Kontext: A new generation of multimodal image generation and editing tools"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"https://kontextflux.com"}},"_tags":["story","author_AllenRen","story_44157477"],"author":"AllenRen","created_at":"2025-06-02T10:48:34Z","created_at_i":1748861314,"num_comments":0,"objectID":"44157477","points":1,"story_id":44157477,"title":"Flux Kontext: A new generation of multimodal image generation and editing tools","updated_at":"2025-06-02T10:52:25Z","url":"https://kontextflux.com"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Equal46"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Try Flux.1 Kontext to Edit Images for Free"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://fluxcontext.org"}},"_tags":["story","author_Equal46","story_44142243"],"author":"Equal46","children":[44142244],"created_at":"2025-05-31T06:21:17Z","created_at_i":1748672477,"num_comments":0,"objectID":"44142243","points":1,"story_id":44142243,"title":"Try Flux.1 Kontext to Edit Images for Free","updated_at":"2025-05-31T06:26:17Z","url":"https://fluxcontext.org"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"howardV"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hey HN!
I've been working on FLUX.1 Kontext, an AI-powered image editing platform that leverages Black Forest Labs' revolutionary FLUX.1 Kontext models.
*What makes it different:*\n- Context-aware editing: The AI understands relationships between objects in your image\n- Natural language prompts: "Change the background to a sunset" actually works as expected\n- Professional-grade results: Built on the same tech stack powering enterprise AI tools\n- Real-time processing: Most edits complete in under 30 seconds
*Technical details:*\n- Frontend: Next.js 14 with TypeScript\n- AI Models: FLUX.1 Kontext Pro/Max from Black Forest Labs\n- Security: Cloudflare Turnstile integration\n- Image processing: Optimized pipeline for various aspect ratios
*Why I built this:*\nExisting AI image editors often miss context or produce unrealistic results. FLUX.1 Kontext models excel at understanding spatial relationships and maintaining consistency across edits.
*Try it free:* [https://konfluxart.com]
Would love feedback from the HN community! What image editing challenges have you faced with current AI tools?
Tech stack details and implementation notes in comments if anyone's interested."},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Show HN: An AI image editor that understands context (FLUX.1 Kontext)"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://konfluxart.com"}},"_tags":["story","author_howardV","story_44137165","show_hn"],"author":"howardV","created_at":"2025-05-30T15:26:40Z","created_at_i":1748618800,"num_comments":0,"objectID":"44137165","points":1,"story_id":44137165,"story_text":"Hey HN!
I've been working on FLUX.1 Kontext, an AI-powered image editing platform that leverages Black Forest Labs' revolutionary FLUX.1 Kontext models.
*What makes it different:*\n- Context-aware editing: The AI understands relationships between objects in your image\n- Natural language prompts: "Change the background to a sunset" actually works as expected\n- Professional-grade results: Built on the same tech stack powering enterprise AI tools\n- Real-time processing: Most edits complete in under 30 seconds
*Technical details:*\n- Frontend: Next.js 14 with TypeScript\n- AI Models: FLUX.1 Kontext Pro/Max from Black Forest Labs\n- Security: Cloudflare Turnstile integration\n- Image processing: Optimized pipeline for various aspect ratios
*Why I built this:*\nExisting AI image editors often miss context or produce unrealistic results. FLUX.1 Kontext models excel at understanding spatial relationships and maintaining consistency across edits.
*Try it free:* [https://konfluxart.com]
Would love feedback from the HN community! What image editing challenges have you faced with current AI tools?
Tech stack details and implementation notes in comments if anyone's interested.","title":"Show HN: An AI image editor that understands context (FLUX.1 Kontext)","updated_at":"2025-05-30T15:29:45Z","url":"https://konfluxart.com"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"gravitywp"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Try Flux.1 Kontext on Fluxk.art"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://fluxk.art"}},"_tags":["story","author_gravitywp","story_44131757"],"author":"gravitywp","created_at":"2025-05-30T00:50:28Z","created_at_i":1748566228,"num_comments":0,"objectID":"44131757","points":1,"story_id":44131757,"title":"Try Flux.1 Kontext on Fluxk.art","updated_at":"2025-05-30T00:53:57Z","url":"https://fluxk.art"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rushingcreek"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hi HN,
Today we\u2019re launching phind.design (https://phind.design), an image editor and design tool that uses 4o and custom models to allow users to generate and edit designs for anything from logos and advertisements to creative website and app designs.
4o is great at producing a first version of an image, but is not capable of editing it without messing up other parts of it. We fix this by running Flux Kontext alongside 4o image gen in the chat, as well as by introducing a precision editor powered by custom models where a user indicates an area to modify and we guarantee that only that area will be modified.
Our precision editor is state-of-the-art on image editing in our tests and allows inserting new additional images into the existing image. The latter allows users to insert a logo, product, or face into an image without messing up other parts of the image, and even fix logos and faces that were messed up by 4o. Text editing with the precision edit model is still a work in progress, and we will fix it with the next iteration of that model. We recommend using the chat for editing text for now.
Example: Insert UT Austin logo into helicopter ad (https://phind.design/edit?chat=cmd27o2n10001l704h6865f3u)
We also always produce multiple variations for image generations and edits, as we think this variety is important for getting exactly what you asked for.
Example: Paul Graham in startup heaven (https://phind.design/edit?chat=cmd23h91c000jky04no5d92uy)
One thing we\u2019re excited about is adding more variation into AI-generated websites, as many website builders all use the same CSS libraries, so many websites end up looking the same. \nWe hope to allow builders and creatives to make truly unique designs in 1/10th the time it currently takes with existing tools.
Example: Make me a popeyes landing page where the eyes are actually popping out (https://phind.design/edit?chat=cmd25imtm0001jr046nsag4lu)\nExample: A train map with sandwich ingredients replacing subway stops. (https://phind.design/edit?chat=cmd23i98c0001ie04l56npyj3)
As engineers who have been frustrated by the time commitment it takes to learn Figma or Photoshop, we hope that phind.design makes it incredibly easy to go from zero to one on your wildest creative ideas.
The editor is far from perfect, particularly when it comes to text. We are working on it and have a new custom precision editing model on the way. In the meantime, we\u2019re excited to hear your comments and feedback!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Phind.design \u2013 Image editor & design tool powered by 4o / custom models"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://phind.design"}},"_tags":["story","author_rushingcreek","story_44650622","show_hn"],"author":"rushingcreek","children":[44653332,44653483,44653564,44654320,44655743,44655822,44656023,44657840,44657918,44658372,44658636,44660549,44661367,44667934,44719370],"created_at":"2025-07-22T17:44:33Z","created_at_i":1753206273,"num_comments":22,"objectID":"44650622","points":90,"story_id":44650622,"story_text":"Hi HN,
Today we\u2019re launching phind.design (https://phind.design), an image editor and design tool that uses 4o and custom models to allow users to generate and edit designs for anything from logos and advertisements to creative website and app designs.
4o is great at producing a first version of an image, but is not capable of editing it without messing up other parts of it. We fix this by running Flux Kontext alongside 4o image gen in the chat, as well as by introducing a precision editor powered by custom models where a user indicates an area to modify and we guarantee that only that area will be modified.
Our precision editor is state-of-the-art on image editing in our tests and allows inserting new additional images into the existing image. The latter allows users to insert a logo, product, or face into an image without messing up other parts of the image, and even fix logos and faces that were messed up by 4o. Text editing with the precision edit model is still a work in progress, and we will fix it with the next iteration of that model. We recommend using the chat for editing text for now.
Example: Insert UT Austin logo into helicopter ad (https://phind.design/edit?chat=cmd27o2n10001l704h6865f3u)
We also always produce multiple variations for image generations and edits, as we think this variety is important for getting exactly what you asked for.
Example: Paul Graham in startup heaven (https://phind.design/edit?chat=cmd23h91c000jky04no5d92uy)
One thing we\u2019re excited about is adding more variation into AI-generated websites, as many website builders all use the same CSS libraries, so many websites end up looking the same. \nWe hope to allow builders and creatives to make truly unique designs in 1/10th the time it currently takes with existing tools.
Example: Make me a popeyes landing page where the eyes are actually popping out (https://phind.design/edit?chat=cmd25imtm0001jr046nsag4lu)\nExample: A train map with sandwich ingredients replacing subway stops. (https://phind.design/edit?chat=cmd23i98c0001ie04l56npyj3)
As engineers who have been frustrated by the time commitment it takes to learn Figma or Photoshop, we hope that phind.design makes it incredibly easy to go from zero to one on your wildest creative ideas.
The editor is far from perfect, particularly when it comes to text. We are working on it and have a new custom precision editing model on the way. In the meantime, we\u2019re excited to hear your comments and feedback!","title":"Show HN: Phind.design \u2013 Image editor & design tool powered by 4o / custom models","updated_at":"2026-05-31T14:24:14Z","url":"https://phind.design"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"chris_sandor"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"My daughter Kate (7 years old) really loves Minecraft! Together, we used several generative AI models (Wan Video, Flux Kontext, and DeepSeek) to create a 1-minute animation based on only 1 input photo of her.
You can read my detailled description of how we made it here: https://drsandor.net/ai/minecraft/ or can directly watch the video on youtube: https://youtu.be/xl8nnnACrFo?si=29wB4dvoIH9JjiLF"},"title":{"matchLevel":"none","matchedWords":[],"value":"Experience Making a 1-minute AI movie with my 7-year old daughter"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://drsandor.net/ai/minecraft/"}},"_tags":["story","author_chris_sandor","story_44375696"],"author":"chris_sandor","children":[44383936,44384024,44384031,44384039,44384496],"created_at":"2025-06-25T10:35:58Z","created_at_i":1750847758,"num_comments":9,"objectID":"44375696","points":22,"story_id":44375696,"story_text":"My daughter Kate (7 years old) really loves Minecraft! Together, we used several generative AI models (Wan Video, Flux Kontext, and DeepSeek) to create a 1-minute animation based on only 1 input photo of her.
You can read my detailled description of how we made it here: https://drsandor.net/ai/minecraft/ or can directly watch the video on youtube: https://youtu.be/xl8nnnACrFo?si=29wB4dvoIH9JjiLF","title":"Experience Making a 1-minute AI movie with my 7-year old daughter","updated_at":"2025-07-13T19:33:04Z","url":"https://drsandor.net/ai/minecraft/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"anjneymidha"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"A Flux Kontext + Mistral experiment. Upload an image, and let the AIs do the rest of the work."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Komposer, AI image editor where the LLM writes the prompts"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.komposer.xyz/"}},"_tags":["story","author_anjneymidha","story_45006777","show_hn"],"author":"anjneymidha","children":[45009442,45010195,45018779,45024376],"created_at":"2025-08-24T19:06:41Z","created_at_i":1756062401,"num_comments":4,"objectID":"45006777","points":12,"story_id":45006777,"story_text":"A Flux Kontext + Mistral experiment. Upload an image, and let the AIs do the rest of the work.","title":"Show HN: Komposer, AI image editor where the LLM writes the prompts","updated_at":"2026-03-05T22:29:42Z","url":"https://www.komposer.xyz/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ringerapps"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hi HN, I\u2019ve been tinkering with this for a while and finally have it in a decent state. It\u2019s a plotter robot that draws caricatures from photos.
I trained a diffusion model (Flux Kontext LoRA) on caricature images, 3D-printed a Sharpie mount for my Ender 3, and hacked together a pipeline that goes photo \u2192 caricature \u2192 G-code. After a lot of trial and error it\u2019s working pretty well, and I put up a little site where you can try it out.
Happy to answer questions or hear any feedback. Thanks!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: I built a robot that draws caricatures with a Sharpie"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.caricature-bot.com/"}},"_tags":["story","author_ringerapps","story_45040211","show_hn"],"author":"ringerapps","children":[45042482],"created_at":"2025-08-27T14:27:48Z","created_at_i":1756304868,"num_comments":2,"objectID":"45040211","points":5,"story_id":45040211,"story_text":"Hi HN, I\u2019ve been tinkering with this for a while and finally have it in a decent state. It\u2019s a plotter robot that draws caricatures from photos.
I trained a diffusion model (Flux Kontext LoRA) on caricature images, 3D-printed a Sharpie mount for my Ender 3, and hacked together a pipeline that goes photo \u2192 caricature \u2192 G-code. After a lot of trial and error it\u2019s working pretty well, and I put up a little site where you can try it out.
Happy to answer questions or hear any feedback. Thanks!","title":"Show HN: I built a robot that draws caricatures with a Sharpie","updated_at":"2026-03-05T22:32:12Z","url":"https://www.caricature-bot.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"rjyoungling"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"typo in the url: https://vibeflirting.com/
Hey HN,\nAfter years of false starts learning to code (pure math background, but got stuck in tutorial hell), I finally shipped something real with Claude Code as my patient tutor.
The problem: Most guys have terrible dating profile photos. Poor lighting, bad angles, unflattering composition. Men generally don't like taking pics, so our photos suck.
My solution: Built a pipeline that takes uploaded photos and enhances them specifically for dating apps by tying AI tools together. Trying different things but had the most success with training flux and using enhancor or kontext for skin texture (and midjourney for minor editing).
Tech stack: Vanilla HTML/CSS/JS frontend, Netlify Functions backend, Cloudinary for storage, Google Sheets as database. No frameworks - kept it simple to actually finish something.
(I also tried Tailwind with CSS but that eventually caused build conflicts, so I removed it and wrote everything in pure CSS. Lesson learned.)
The learning curve was steep, but Claude Code answered literally hundreds of basic questions. I used todo folders to save context between sessions and screenshot folders for debugging help. Some other things too but with the pace of innovation, many of those are already outdated.
Took about two months to get it all working. Mostly just happy to have shipped something and "vibelearned." Hoping that by shipping and learning incrementally, I can tackle more complex projects.
All advice for this new indie hacker/newbie coder is welcome.
RJ"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Vibeflirting turns bad selfies into good dating pics for Tinder"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://vwww.vibeflirting.com/"}},"_tags":["story","author_rjyoungling","story_44390620","show_hn"],"author":"rjyoungling","children":[44390727],"created_at":"2025-06-26T19:46:26Z","created_at_i":1750967186,"num_comments":3,"objectID":"44390620","points":3,"story_id":44390620,"story_text":"typo in the url: https://vibeflirting.com/
Hey HN,\nAfter years of false starts learning to code (pure math background, but got stuck in tutorial hell), I finally shipped something real with Claude Code as my patient tutor.
The problem: Most guys have terrible dating profile photos. Poor lighting, bad angles, unflattering composition. Men generally don't like taking pics, so our photos suck.
My solution: Built a pipeline that takes uploaded photos and enhances them specifically for dating apps by tying AI tools together. Trying different things but had the most success with training flux and using enhancor or kontext for skin texture (and midjourney for minor editing).
Tech stack: Vanilla HTML/CSS/JS frontend, Netlify Functions backend, Cloudinary for storage, Google Sheets as database. No frameworks - kept it simple to actually finish something.
(I also tried Tailwind with CSS but that eventually caused build conflicts, so I removed it and wrote everything in pure CSS. Lesson learned.)
The learning curve was steep, but Claude Code answered literally hundreds of basic questions. I used todo folders to save context between sessions and screenshot folders for debugging help. Some other things too but with the pace of innovation, many of those are already outdated.
Took about two months to get it all working. Mostly just happy to have shipped something and "vibelearned." Hoping that by shipping and learning incrementally, I can tackle more complex projects.
All advice for this new indie hacker/newbie coder is welcome.
RJ","title":"Show HN: Vibeflirting turns bad selfies into good dating pics for Tinder","updated_at":"2025-10-06T20:22:25Z","url":"https://vwww.vibeflirting.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"lschneider"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hi HN,
YCW24 company here, we just open-sourced an AI image detection model that beats the SOTA commercial detectors. AI-generated images/videos have become incredibly good in the last few months and are flooding the internet; being able to detect them reliably gives some power back to consumers and companies that care about high-quality, genuine content.
Detecting AI-generated images is a very hard problem: there are many different techniques to generate images; there is image compression, noise, and other distortions that destroy generator artifacts; there's android phones applying auto-correction to images; etc. And none of the detectors we tried (sightengine.com, decopy.ai, etc.) works reliably even for basic examples (try it out with pirate Rick Astley made with Flux Kontext: https://imgur.com/a/iL3paE8).
We released two models, the full version (~600M params) and a smaller version (~20M params) that can even run in your browser on mobile (see demo)! We've also put up code for running things locally or via an API (free but rate-limited) using javascript/node and python code.
-------- Details --------
The full model was trained on 1M+ images that were scraped off the internet and the small model is a distillation. We're actively working on extending the dataset and further improving the models.
Classification accuracy: sightengine.com seems to be the best commercial solution out there, as confirmed by this (https://arxiv.org/pdf/2404.14581) paper, which they also cite on their website. Of course, they cherry-picked the results and claim 98.3% accuracy while only achieving (still impressive) ~82.8% over the full dataset. I've downloaded the dataset used in the paper and tested my models against it. The code for running the tests as well as a usable version of the dataset (the original was a big pain to download from OneDrive) are included in the repo code.
Here are our benchmark results for comparison:\nTotal samples: 144,088\nReal images: 17,044\nSynthetic images: 127,044\nAverage precision: 0.991
[threshold: 0.5] \n Total accuracy: 0.864
PER-CATEGORY ACCURACY: \n Real 0.875 (17,044 samples) \n DALL-E_T2I 0.982 (16,110 samples) \n DreamStudio_T2I 0.968 (16,278 samples) \n Midjourney_T2I 0.961 (16,148 samples) \n StarryAI_T2I 0.847 (13,515 samples) \n DALL-E_IT2I 0.774 (16,665 samples) \n DreamStudio_IT2I 0.666 (16,139 samples) \n Midjourney_IT2I 0.897 (15,371 samples) \n StarryAI_IT2I 0.805 (16,818 samples)
[threshold: 0.65] \n Total accuracy: 0.827
PER-CATEGORY ACCURACY: \n Real 0.914 (17,044 samples) \n DALL-E_T2I 0.971 (16,110 samples) \n DreamStudio_T2I 0.949 (16,278 samples) \n Midjourney_T2I 0.940 (16,148 samples) \n StarryAI_T2I 0.803 (13,515 samples) \n DALL-E_IT2I 0.698 (16,665 samples) \n DreamStudio_IT2I 0.576 (16,139 samples) \n Midjourney_IT2I 0.845 (15,371 samples) \n StarryAI_IT2I 0.743 (16,818 samples)"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Open-source AI image/deepfake detection that actually works"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://www.nonescape.com/"}},"_tags":["story","author_lschneider","story_44797022","show_hn"],"author":"lschneider","created_at":"2025-08-05T12:03:45Z","created_at_i":1754395425,"num_comments":0,"objectID":"44797022","points":3,"story_id":44797022,"story_text":"Hi HN,
YCW24 company here, we just open-sourced an AI image detection model that beats the SOTA commercial detectors. AI-generated images/videos have become incredibly good in the last few months and are flooding the internet; being able to detect them reliably gives some power back to consumers and companies that care about high-quality, genuine content.
Detecting AI-generated images is a very hard problem: there are many different techniques to generate images; there is image compression, noise, and other distortions that destroy generator artifacts; there's android phones applying auto-correction to images; etc. And none of the detectors we tried (sightengine.com, decopy.ai, etc.) works reliably even for basic examples (try it out with pirate Rick Astley made with Flux Kontext: https://imgur.com/a/iL3paE8).
We released two models, the full version (~600M params) and a smaller version (~20M params) that can even run in your browser on mobile (see demo)! We've also put up code for running things locally or via an API (free but rate-limited) using javascript/node and python code.
-------- Details --------
The full model was trained on 1M+ images that were scraped off the internet and the small model is a distillation. We're actively working on extending the dataset and further improving the models.
Classification accuracy: sightengine.com seems to be the best commercial solution out there, as confirmed by this (https://arxiv.org/pdf/2404.14581) paper, which they also cite on their website. Of course, they cherry-picked the results and claim 98.3% accuracy while only achieving (still impressive) ~82.8% over the full dataset. I've downloaded the dataset used in the paper and tested my models against it. The code for running the tests as well as a usable version of the dataset (the original was a big pain to download from OneDrive) are included in the repo code.
Here are our benchmark results for comparison:\nTotal samples: 144,088\nReal images: 17,044\nSynthetic images: 127,044\nAverage precision: 0.991
[threshold: 0.5] \n Total accuracy: 0.864
PER-CATEGORY ACCURACY: \n Real 0.875 (17,044 samples) \n DALL-E_T2I 0.982 (16,110 samples) \n DreamStudio_T2I 0.968 (16,278 samples) \n Midjourney_T2I 0.961 (16,148 samples) \n StarryAI_T2I 0.847 (13,515 samples) \n DALL-E_IT2I 0.774 (16,665 samples) \n DreamStudio_IT2I 0.666 (16,139 samples) \n Midjourney_IT2I 0.897 (15,371 samples) \n StarryAI_IT2I 0.805 (16,818 samples)
[threshold: 0.65] \n Total accuracy: 0.827
PER-CATEGORY ACCURACY: \n Real 0.914 (17,044 samples) \n DALL-E_T2I 0.971 (16,110 samples) \n DreamStudio_T2I 0.949 (16,278 samples) \n Midjourney_T2I 0.940 (16,148 samples) \n StarryAI_T2I 0.803 (13,515 samples) \n DALL-E_IT2I 0.698 (16,665 samples) \n DreamStudio_IT2I 0.576 (16,139 samples) \n Midjourney_IT2I 0.845 (15,371 samples) \n StarryAI_IT2I 0.743 (16,818 samples)","title":"Show HN: Open-source AI image/deepfake detection that actually works","updated_at":"2025-08-11T16:17:59Z","url":"https://www.nonescape.com/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"Cilvic"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["kontext"],"value":"Hey HN, \nI kept pasting screenshots into cursor & chatGPT and it inspired me how much relevant context my screen can provide to multi-modal LLMs.
To make it easier to grab the user context inside of our own apps, together with Fernando and Claire we\u2019ve built an SDK to perform screen capture & OCR. It\u2019s written in Rust and runs on Mac, Windows & Linux.
We have published bindings for Node.js, to try the Node sample:
1. Git clone kontext21/k21-node-sample\n2. Npm install\n3. Npm run start
This will capture what\u2019s on your screen and OCR the text to console.
Everything is local and there is no telemetry. It\u2019s source-available so you can dig deeper if you are curious .
Any feedback is welcome as we are getting started with this
- https://github.com/kontext21/k21-node-sample/\n- https://github.com/kontext21/k21\n- https://docs.kontext21.com/"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Cross-platform screen capture & OCR SDK, Rust with Node bindings"}},"_tags":["story","author_Cilvic","story_43692293","show_hn"],"author":"Cilvic","children":[43692321],"created_at":"2025-04-15T13:20:21Z","created_at_i":1744723221,"num_comments":1,"objectID":"43692293","points":2,"story_id":43692293,"story_text":"Hey HN, \nI kept pasting screenshots into cursor & chatGPT and it inspired me how much relevant context my screen can provide to multi-modal LLMs.
To make it easier to grab the user context inside of our own apps, together with Fernando and Claire we\u2019ve built an SDK to perform screen capture & OCR. It\u2019s written in Rust and runs on Mac, Windows & Linux.
We have published bindings for Node.js, to try the Node sample:
1. Git clone kontext21/k21-node-sample\n2. Npm install\n3. Npm run start
This will capture what\u2019s on your screen and OCR the text to console.
Everything is local and there is no telemetry. It\u2019s source-available so you can dig deeper if you are curious .
Any feedback is welcome as we are getting started with this
- https://github.com/kontext21/k21-node-sample/\n- https://github.com/kontext21/k21\n- https://docs.kontext21.com/","title":"Show HN: Cross-platform screen capture & OCR SDK, Rust with Node bindings","updated_at":"2025-04-15T13:34:45Z"}],"hitsPerPage":50,"nbHits":16351,"nbPages":20,"page":0,"params":"query=Kontext&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":10,"processingTimingsMS":{"_request":{"roundTrip":17},"afterFetch":{"format":{"highlighting":2,"total":2},"merge":{"mergeLoop":{"prepareNextHit":1,"total":1},"total":1},"total":1},"fetch":{"query":3,"scanning":4,"total":8},"total":11},"query":"Kontext","serverTimeMS":13}