Chris Messina

Generated Photos API — Generate worry-free, diverse models on-demand using AI

Featured
51
The most practical way to get high-quality generated faces for commercial content.
• Fully tagged and searchable faces
• Forget about sourcing, model releases, and hassles!
• AI-flaw detection for incredible consistency
• Trained on licensed, exclusive data

Add a comment

Replies
Best
Ivan Braun
When we launched the 100,000 Faces tech demo a few months ago we also teased our main work, creating a versatile API that serves AI-generated content. Today we are happy to take the wraps off of our v1 API. Now you can easily integrate permissive AI-generated headshot photos into any project. This new API offers fully tagged and searchable access to faces across age ranges, ethnicities, and physical attributes. We are constantly improving our output results and have developed a brand new AI system to make sure faces that seem too ‘off’ never make it to production. This new process delivers class-leading consistency for AI developed imagery. Get started for free today! Personal use is available for free with attribution and commercial licenses are now available as well. We are already creating new products with this API internally and are very excited to see what the community produces. Interest for headshot images has already come in from research projects, dementia testing applications, 3D facial animation programs, advertisers, and many more creative cases that have been left behind by traditional stock media. If you have custom needs or a showcase project you want to use AI-generated images on, we would love to hear about it. PS: Also launching today is our main photo platform! Filter, sort, download, and vote on 100k+ pre-created headshot images; all powered by the Generated Photos API.
Ivan Braun
PS ProductHunt rules prohibit us releasing it yet, but here's more fun: https://generated.photos/faces The new better faces with filters and my favorite feature: similar faces
Marina Yalanska
Cool to see such a progress for quite a short time after the first launch! Kudos to the team
Ivan Braun
@myalanska many many thanks!
Tyler Lastovich
Media coverage for AI typically focuses on deep fakes, but we chose to see things on the brighter side. AI-based media has many wholesome uses such as for people who are trying to protect their anonymity, avoid embarrassing situations, or reach new audience members through image tailoring. With this release we hope to start moving generated content out of 'vaporware' and into the hands of real users. What makes Generated Photos special? - We have developed the safest AI production process, which involved taking tens of thousands of model-released images in our own studio. - Our model images are made from scratch by our machine learning systems, meaning they carry no additional usage restrictions such as copyrights, or sensitive-use clauses. - Our images feature incredible detail and consistency compared to other generative offerings available. - We offer the ability to directly integrate our faces into nearly any app or website on-demand and at scale.
Ivan Braun
@tylerlastovich @alxcnwy that was my understanding too at first: scrape whatever photos you need and generate whatever you need. Apparently, it's not that simple; you can't using people's personal data even if you don't publish it openly. Facebook got sued for much less: https://techcrunch.com/2019/10/1... I'm not a lawyer myself and can't give legal advice, but here's what I've learned from the talks with our attorney: if you're a large business and use scraped data, some of these scraped models will actively seek for another consent you miss or another law you broke.
Alex Conway
Well done Ivan & team! This is a really interesting space, congratulations on your API launch and on shipping so quickly! 🙌 I’m keen to see what products you’re creating internally… Do you only do headshots or full-body generated images? BTW clicking "refuse" on your cookies popup redirects me here for some reason: https://img.icons8.com/
Tyler Lastovich
@alxcnwy Thanks Alex! We are definitely working with more than just headshots and are actively capturing data in our studio well beyond just faces. Faces were a logical space to start with for us, as there are a fewer edge cases to handle for ML generation (clothing is hard because it comes in so many styles). In the meantime, we have face transfer to existing images fully working and that will be one of the next releases we push out. And thanks for the bug report, we are fixing a few little things right now. :)
Alex Conway
@tylerlastovich Awesome, that sounds like a solid start! I'd love to see you guys productize and refine some of the mind-blowing GAN "try-on clothes" work e.g. https://github.com/sergeywong/cp...
Ivan Braun
@tylerlastovich @alxcnwy what are the applications of it? * Making the product shots for ecommerce? * Virtual dressing room?
Maxime Patte
@tylerlastovich @alxcnwy that's awesome! Do you know more work on fashion like that?
Ivan Braun
@tylerlastovich @alxcnwy @maxime_patte vue.ai coven.ai facebook's fashion++ etc.
Bruce
Hey, tried visiting your site and you got a 502 Bad Gateway.
Ivan Braun
@bcye @looksawesome how about now? PS Sorry for a hickup. Added a CNAME for our mailing; one thing lead the another...
Ivan Braun
@bcye @looksawesome thanks for letting know! DNS cache it is. I was worried, pushing the team for some weird fixes like nginx proxy from the old server to the new one.
Sasha Yakovlev
Disgusting! Started to use generated.photos few hours ago and can't stop. Still trying to build ideal human...
Ann Narkevich
Congrats on the launch guys! Best of luck 🙌
Ivan Braun
@anna_narkevich many thanks! many many 🖤
Cedrik Dudek
404 when trying to reach https://generated.photos/api :(
Ivan Braun
@cedrik_dudek sorry for that! You've got DNS cached, I have it myself on one of my phones. It must be over now. If not, it depends. 1) If you're involved enough, you may flush your cache; it must resolve to 35.232.31.120 2) If not, just hate me 😌 I've earned it.
Ivan Braun
@cedrik_dudek pheww :) We had a meeting here, thinking up different weird fixes: an nginx proxy on the old host, accessing by ip, etc.
Jaeseok An
What are the reasons asian male faces are super rare? For example, follow link shows 99%+ of female faces only unlike White, Latino, and Black. https://generated.photos/faces/a...
Ivan Braun
@jaeseok_an we can plenty asian males: https://generated.photos/faces/a... But you're right, the gallery tends to bias toward certain characteristics (not only ethnic). I'm not sure about it, but this could be either a generation, filtering, or output issue. Or, some combination of them. There are two ways to fix it: 1. Balance the training set — we're busy doing it 2. Balance the output on the website, i.e. artificially mix the images to make it diverse. We're going to do it in our next version Btw, how do you think we should show the output? 1. For users in China, should the majority be asians? 2. In general, should we mind the ethnic profile of planet earth? 3. Of a country where a visitor is from? 4. Of a city where a visitor is from?
Jaeseok An
@visualpharm Thanks for your answer! Regarding your questions 1. For users in China, should the majority be asians? -> I don't think it's a necessary feature to show different results based on user's profile. Also I wouldn't assume people would use this product to find someone like them (eg, race / gender). 2, 3, 4 -> I don't think they are necessary.
Ivan Braun
@jaeseok_an so just equal proportions of all features?
Jaeseok An
@visualpharm Equal proportions of all features -> what do you mean by features here? race, gender, country, city of the face models for training your product?
Ivan Braun
@jaeseok_an yes. If there should be enough smiling, surprised, angry, disgusted, tortured, dreaming, sleeping etc people. If there should be same proportion of blondes, even though it's a recessive gene? My point is balancing the set is hard. Not impossible to approximate, but hard. A "normal" distribution depends highly on what we consider human, not equal parts of everything.
Sharath Prabhal
Congrats on the launch! I've been waiting for the API ever since I first got wind of Generated Photos. One question - is there an easy/suggested way to use a custom background for the images? The plain/white background makes it look very formal.
Ivan Braun
@sharathprabhal yep. We have masked all of them and can deliver them with a transparent background.
Vic Koch
Congratulations on this release! We used generated avatars for profiles, cause not many people want to show real face. I think it’s very useful to keep privacy to financial institution inspectors and security officers.
jafbm🧢🩺🎻🎸🚀🤖
They aren't free. What's the point of paying for something when you can get something free?
Ivan Braun
@jafbm57 You can download any pictures for free. Try clicking Download here: https://generated.photos/face/jo... However, I feel that you see a free alternative for something we charge for. Where should we tweak our freemium model?
Rudrakshya Barman
When I choose Gender=Male, Age=Adult, Ethnicity=White, Eye color=brown, Hair color=black, hair length=medium, emotion=joy, Its shows 1 Female image too.
Ivan Braun
@rudrakshya_barman quite a test case! Will check it out. These filters are not 100% bulletproof though. They're way better than mturk crowd, and way better than a human, but still have an error margin of around 8%.
Colin Frederick
This is amazing! Is there an ability to do face replacements in video content?
Ivan Braun
@producthunt_film not yet! We're couple challenges away of it, but that's gonna be an awesome feature.
Colin Frederick
@visualpharm Sweet! There are so many projects I'd like to use that for. Is it possible to make multiple photos of the same fake person? Like, with different angles? If so, I could use photogrammetry to potentially make a 3d model with the still images. Is there anyway to play with the lighting too? This is so exciting I can't stop thinking of all the possibilities
Ivan Cucer
@visualpharm lol it'd be solid to have a sketch plugin for designers 🤪
Amir Kerawala
Great product, I can certain use it for my business.
Andre Ruiz
Good application of AI.
Konstantin Valter
It will be also useful if you add feature to generate several different photos of the SAME person
Plamen Donev
It would be good to introduce control for making the faces less "supermodel"-ish. They become more real looking when they have flaws, like slight asymmetry, moles etc. Also - facial proportions, in order to make someone less pretty.
Tony OHagan
Wonderful!! I really like the fact that you've added racial diversity. The faces are still very US / European in gesture and style and there's still lots to do to fully support an international audience. For example ... think Middle-east, India/Pakistan ... for which you'd need to include religious head coverings or facial changes (think Islamic, Jewish, Sikh). For South-Pacific islands .. Maouri, Fijian, Samoan, New Guinea. For Australians ... Country vs City (think freckles, different facial gestures). Great start. Tons still to do to cover faces.