Subscribe
Sign in
100,000 Faces — Free resource of 100k diverse faces generated by AI
We are happy to share with you a massive free resource of 100k faces, generated from scratch using AI.
This project showcases a glimpse into the technology we are building to create media on-demand.
134
Replies
Best
@_jacksmith i think they addressed this in the initial comment but not sure if that's a full enough explanation for what youre looking for
@anderson760 yeah, it doesn't answer it. as "These images were produced completely by our artificial intelligence — none of these people are real!" may be true, but they highly likely would have been based on real underlying people's photos. else if the AI didn't have anything to compare against, how would it know that it's photos were realistic? you need underlying examples in order to train an AI model.
@anderson760 @_jacksmith great point. We've run a photo lab for the last 3 years. We shot all the photos that we use. All of them are model-released.
The next question is: what if some of the generated photos reminds some live person? I don't know an answer here. I doubt anybody has. We have the whole new legal system here.
@anderson760 @visualpharm cool. that addresses the point that I had then (the fact you shot all of the photos that you use)
Hey @ikalimullin @pavel_malay @zzzhabinskiy @visualpharm
I made a simple API.
This URL gives you a random image
https://100k-faces.glitch.me/ran...
And this one just a random image URL as JSON
https://100k-faces.glitch.me/ran...
Here's the demo
https://100k-faces.glitch.me/
Here's the source code
https://github.com/ozgrozer/100k...
And here's the Product Hunt example
@ikalimullin @pavel_malay @zzzhabinskiy @ozgrozer this is awesome! I'm adding it to the top bar with all the news and publications.
Yes. This is exactly what I’ve been looking for 👏
Wanna share with us what are the next projects/tools you’re working on?
@hypervillain an AI model agency. You'd be able to
1. Choose an AI model
2. Send us your merchandise: gadgets, clothing, jewelry
3. Receive the photos of it.
@hypervillain @visualpharm Let me know when you are looking for partners on this. I have several apparel companies I'm involved with. Have been looking for something like this 👍🏼
@hypervillain @visualpharm @ryanwmark Sounds a lot like www.artificialtalent.co
I'm the co-founder and a former Ford Model - your problem will be controlling for full body posing which cant be done at present with AI alone. You could simply put the AI face on a real model but that'd defeat most of the cost savings youre trying to keep by avoiding a human asset. AI is excellent for generating novel faces on demand however, only a part of the equation for fashion modeling.
Let me know if you want to chat about it, would love to involve you.
@hypervillain @visualpharm @ryanwmark @b_a_randall Ford Models was founded in 1946 by Eileen Ford and her husband Gerard W. Ford ?
I have seen so many AI-generated faces. They look beautiful and seem natural. What are the practical uses of these AI-generated faces?
Hey PH, we are really excited to give you a look at our upcoming tech.
We have put together a free resource of 100k faces for you to use however you wish. But these aren't just normal faces. These images were produced completely by our artificial intelligence — none of these people are real!
We are currently working hard to train and refine our generative models toward our ultimate goal: creating a simple API that can produce infinite diversity. We are iterating fast, but things aren’t perfect quite yet. When you see a face that is a bit ‘off’, just give it some slack. 😅
We also know that the future can be unnerving at times. That is why we went above and beyond to assure personal safety in our systems. All of our training data was shot in-house and is fully model released. No images were pulled from stock media or scraped from the internet. This requires thousands of hours of labor, but in the end we know it will be worth it!
These faces feature:
- Consistent lighting
- Consistent sizing
- Range of angles, positions
- Range of facial expressions
- Wide variety of ethnicities
- All ages
- All face shapes
Have your next mockup, presentation or website looking great with these fresh faces.
Everything is free for non-commercial use as long as you include link attribution back to https://generated.photos, so that more people can find us.
Have fun and let us know what you make!
@zzzhabinskiy hey hey hey. This is awesome. Finally won't need that copyright stuff. Very cool and congrats to you. How has your launch been going?
Hey @drew_dunn1, thanks a lot! The launch has just started so it seems to be a bit early to strike any balance.
@zzzhabinskiy can I download your photos and use it on my app to let people get a taste of this? I don't mind crediting it to your team. Please let me know. Thanks.
@gainingam_phaomei It'd be glad to the team! 🙌 Sure you could use it, but with a backlink.
I already have asked about links in different thread, so here you go: "If you plan to use photo on your website, set a link to Generated Photos on all pages where you use our content. If you use it on most pages, a link in your footer is fine. Desktop and Mobile apps should have a link in the About dialog or Settings. Also, please credit our work in your App Store or Google Play description (something like "Photos by Generated Photos" is fine)."
Great stuff! Couple of questions:
1. You mention a backlink isn't required, but I can't find specification of where/how to link back. Main website? Any place in particular you need me to post this?
2. You showcase an API that implies all images are tagged. Do you have these exposed in a public library too?
3. Do you have a version of the images without background? Gonna save huge chucks of time not having too crop it myself;)
@olivierjanssen thanks!
1. Backlink is required, so if you plan to use photo on your website, set a link to Generated Photos on all pages where you use our content. If you use it on most pages, a link in your footer is fine. Desktop and Mobile apps should have a link in the About dialog or Settings. Also, please credit our work in your App Store or Google Play description (something like "Photos by Generated Photos" is fine).
2. Not yet, but we're about to launch it soon. Can't wait to share it though :)
3. Unfortunately not yet. There is a huge room for improvements, so we probably will make it in further updates.
@zzzhabinskiy: Great work! You are onto something amazing. A couple of recommendations (if you can implement it)
1. I don't see "Indians" as meta data anywhere. (even though there is "Asians"). I hope you implement that too because that's roughly 1.3 Billion people.
2. Any plans for animated characters? See a lot of use case in Anime/Manga world.
Great work again! Will be following your work closely
@pratik_shetty thank you for such a detailed feedback!
1. We didn't launch our API yet, so there is only examples of API requests we have on promo page. But we totally will implement that meta filter too.
2. It's my dream too, so I hope we could make it in further updates.
@zzzhabinskiy Cheers! I see I made a typo in my question about the backlinksxD (isn't = is). Clear answer, but will be a challenging hurdle when using Generated Photos in client projects (websites). Could you consider crediting in one place in stead of every place?
I can help you with 2. if you want, but I take it that you're about to finish?
About 3.: Does your algo create the background separately or does the background result from merging images?
And finally, a suggestion for a feature: It would be a to select a Generated Photo and use the Unsplash API to combine it with a background
Super cool product - looking forward to hearing how customer usage patterns develop as I imagine they will be genuinely fascinating! Do you think that any of these faces could become an AI generated avatar celebrity or do you think people will generate a new face for every use case?
@liam_roberts that's what we're thinking about too.
I predict:
* New influencers super-optimized for the purpose. Remember Moliere comedy? A villain is super-super-villain. A fitness trainer is super-super-fit. Even his face is super-fit, and when he talks, his face muscles move like crazy.
* A new look of the crowd. We'll have large audiences composed of various people that are good utterly diverse, good looking, and suit the story.
* Modeling is dead as we know it. It's Epstein-prone, discriminatory, and inefficient.
* Creative photography is as we know it.
* The biggest earthquake going to happen in movie making.
Wow, real great sources. I don’t know what to do with them, but at least it can be used in product advertising.
What is your thoughts?
@dimpiax haha we were thinking about it too. UI faces maybe, for the mockups.
What we're going to do next is a model agency. You choose a model, we create the photos with her.
@dimpiax @visualpharm curious how does this work at scale? Are you ultimately having people send you products and then you have certain models pose with those specific products every time? Then that person can be turned into multiple different people/races but with the same backdrop environment? So you ultimately eliminate the need to do a photoshoot with multiple people and can instead do it with just one?
@visualpharm @anderson760 looks like, probably needs of photoshoot and models will be absent, cause of economy and legal contracts. So it will be a boom from one side, and end of *real* not famous users advertisement insta-era from another.
From other perspective, I've got the idea, these resources can be used for potential meeting of finding your "part". It's like, you have an application, then you choose preferable face and parameters (train a model for ML), based on this information, AI sharps your range, and then application, in automatic way, finds variants of persons.
it is definitely an awesome shortcut for portrait usage. Thumb up! One thing to be enhanced maybe is the possibility to expand the head area to bust or even body. And it would be cool that each generated portrait has a unique ID number for easy sharing and bookmarking.
Love it! Some of them generated for a little too long though - certainly seeing some growths and tumours. https://drive.google.com/open?id...
@bradular this is horrible! We're segmenting them to filter out the broken ones.
StyleGAN is fun. I was playing around with it a bit. For example here the images were created using StyleGAN from RunwayML’s desktop client on their “remote GPU” (your repository could have saved me that step :):) ). SqirlZ-Morph was used for the face morph transitions. At the end there is a manual mix of 3 different style transfer images. The facial animation was made with MugLife. Video was edited with Videoleap, Audio was created with SampleBot.
https://youtu.be/IOaHu5ggAWY
Fantastic legal and free (!) resource to build personas, show context better and to illustrate content. Thanks a lot for that.
Three little 'but's, @ :
1. There are only faces of 'beautiful' people, mostly young. Would be better with more diversity: Tubby and fat people, very slim people, people with pimples, bald guys and ladies, people with a lot of makeup, people without any makeup --- you know what I mean. The whole sprectre of humankind
2. I wish there were more people showing strong emotions: Pure happiness, grief, anger, love, thoughtfulness, enviousness, surprise, curiosity....
3. But this is a very tiny 'but': There are many faces where something is wrong with there eyes. But most of them are brilliant. Thanks again!
@zzzhabinskiy The adults look great, you need a LOT more diversity, and the children look...too old. Something about the eyes and nose? A lot of research shows that mammals have a lot of neurons that recognize how old young mammals are, both their own and other species.
@zzzhabinskiy @davemill I agree. We need more training data. For that, we need more children in our studio.
Children are harder to deal and get tired quickly. But we need them. Challenge accepted.
@yash haha, that's what our first target. The industry that robs photographers and washes away 70% to compete on a saturated market — it should go as we know it.
we have a plan Developing an AI-based fashion recommendation AR app. we are figure out those technologies: "Flutter for app development and Tensorflow for image processing & model generation".
can any one suggest instead of this anything else?
@jovisjoseph Some uncanny valley shit here - https://drive.google.com/file/d/...
I'm here if you need any help for building Adobe XD plugin or even collaboration with UI Faces :)
Hustle X