[HN Gopher] Twogether AI - Multi-Person Photo Generation API
___________________________________________________________________
Twogether AI - Multi-Person Photo Generation API
Hey everyone, at Magicflow (YC W23) we're helping our customers run
AI image generation in production, enabling them to produce high-
quality photos at scale. We are launching a scalable API today
that makes it possible to create multi-person portrait photos:
which means the ability to create real-looking photos of any two
persons interacting with each other in some way only by providing a
prompt and the person's pose. Generating this kind of photo
requires a deep understanding of the AI ecosystem, a knowledge gap
many companies face. In order to make the photos look real with
high consistency and for a low cost, chaining of many models is
required, and an excellent understanding of how to tweak with the
various params of each one. We also handle the infrastructure
required to generate the photos, which can be a challenge when
dealt with alone, especially for companies with a small backend
team (we can scale to thousands of requests per day and generate
100 photos in about 3 minutes). Our customers today use this
technology for the following use cases: creating new photo albums
from old-scanned albums, providing personalized content for user
acquisition campaigns, enabling new kinds of experiences in
physical venues, and creating humorous photos with celebrities.
There is a significant tradeoff between creating a robust
abstraction layer on top of Stable Diffusion capabilities and
providing customers with more control over various options. The API
currently allows you to manipulate the following parameters: the
pose of the couple (hugging, taking a selfie, etc.), their facial
expressions, the style of the photo (realistic, cartoon, painting,
etc.), as well as the location, theme, and outfits (e.g., ski
vacation, on the beach) We created a free demo app for you to view
examples and try live: https://twogether.ai?source=hn (no user or
payment needed). For full API access, contact me at
yardenst@magicflow.ai. We can typically set you up within a day,
but an onboarding session is required to ensure responsible API
usage.
Author : yardenst
Score : 24 points
Date : 2023-11-13 15:07 UTC (7 hours ago)
(HTM) web link (twogether.ai)
(TXT) w3m dump (twogether.ai)
| nittanymount wrote:
| will you enhance this site to allow uploading 2 persons photos to
| make a picture together ? thanks
| yardenst wrote:
| It is not something we allow now for various reasons in the
| demo website, but please contact me at yardenst@magicflow.ai
| and we will make what you need.
|
| Using the API it's possible to choose any 2 persons.
| mtlmtlmtlmtl wrote:
| Sounds like an ideal tool for misinformation and smear jobs.
|
| "Look, here's person X with Jeffrey Epstein", etc.
| yardenst wrote:
| For this reason we're currently screening all new customers and
| monitoring the usage
| bigmattystyles wrote:
| so I had this thought - we should flood the zone with fakes.
| All out. Systems like Clearview and pimeyes, and intelligence
| agencies would have a lot more noise to deal with as fake
| detection is expensive per photo and by definition, always
| lagging behind the cutting edge fake generator.
|
| I would also be surprised if governments aren't also sewing
| distrust with their enemies by generating fakes of government
| officials in past random photos of graduating classes of
| military academies of foreign governments, just to sew chaos.
| And why stop there - are you a middle-management cartel member
| with low prospects for advancement? Find a higher ranking
| cartel member you want to get rid of, start generating their
| face in older real photos of FBI graduating classes. Who knows,
| a position might open.
|
| It's a new world.
| JohnFen wrote:
| > so I had this thought - we should flood the zone with
| fakes.
|
| This seems to me like the worst possible thing to do. We
| already have a serious problem being able to determine what's
| true and what's not. A response like this would make the
| problem so much worse that it could pose an actual societal
| threat.
| bigmattystyles wrote:
| That's my point, we might regain an ounce of privacy in
| public.
| JohnFen wrote:
| Oops, my bad. I got Poe'd.
| lancesells wrote:
| I'm not sure where you live but you want intelligence
| agencies to be less effective? There are numerous complaints
| to make against intelligence agencies but I don't know if I
| want them to be worse. I just want them to operate within the
| law and be morally sound.
| comment_ran wrote:
| Dude, this is crazy.
| yardenst wrote:
| Stable Diffusion's eco system is pretty wild!
| seemaze wrote:
| Any website that hijacks my browsers back button is a hard pass.
| yardenst wrote:
| Sorry, we try to move fast, we'll fix it in our next release
| xydac wrote:
| Soon Samsung would be like able to show moon and Saturn twogether
| in their photos
| swatcoder wrote:
| As a funded corporation pursuing sales and not just a casual
| hobbyist doing a tech demo, I have to ask:
|
| How did you get a license to use the image of these celebrities,
| let alone in commercialized deepfakes? Because that's kind of a
| hot topic right now and you're putting a giant target on your
| investors' assets if you didn't bother.
| yardenst wrote:
| Great point, when using the API the photos used to train and
| generate photos are provided by the customer, and they must
| have the correct permissions to use them
| Wistar wrote:
| That misses the point that was being asked which was about
| the celebrity photos and not the user photos.
| yardenst wrote:
| By customer I mean our own customer (a business) and not
| the end user
| swatcoder wrote:
| Am I correct to read that as "we didn't"?
|
| If you make a tool that makes copyright infringement easy,
| _and you demonstrate infringing use in order to market it_ ,
| you're really asking for trouble.
|
| If you care about your startup and its survival, I strongly
| suggest you focus your marketing on legitimate B2B use cases
| and avoid showing off how great it is at making deepfakes.
| It's one thing to turn a blind eye to misuse of your
| technology and a whole different thing to explicitly promote
| misuse as a product feature.
| p2hari wrote:
| Seriously, Gandhi had given permission to generate the photos
| on the website :). On one hand, openAI trains its data on
| copyrighted material. Now YC has funded this startup. A
| significant number of readers have rightly pointed out
| Copyright issues and the founders are saying it is up to
| customers and all those answers have been rightfully
| downvoted. I don't know what else to say here. How could this
| point be overlooked? If I had permission from the celebrity I
| would have taken a photo too. Then what is the need for this.
| wferrell wrote:
| This is so fun. I'd love to integrate this into a holiday card
| app :)
| cj wrote:
| > Our customers today use this technology for the following use
| cases: creating new photo albums from old-scanned albums,
| providing personalized content for user acquisition campaigns,
| enabling new kinds of experiences in physical venues, and
| creating humorous photos with celebrities.
|
| I'm genuinely curious what you see the use case being 5-10 years
| from now. Or in the short term, which use case do you think is
| most compelling from a business perspective? Or is this just an
| entry point to much different uses of AI+Images in the future?
|
| Do you see this as a B2B play or a B2C play?
|
| Lots of AI startups doing various things with AI models, but I
| always have trouble seeing the business model vision. (Or maybe
| that doesn't matter, yet)
| yardenst wrote:
| As a young startup, this is something we are also trying to
| figure out. Currently, it's evident that this technology
| generates the most value in the B2C sector. However we also
| observe a robust B2B demand for other AI models (e-commerce)
| warthog wrote:
| why not make this a consumer app like can of soup? Bet it would
| blow up and make a mess out of BeReal feed
| yardenst wrote:
| Now we are focusing on serving other businesses, if you have a
| good consumer app idea please let's talk over email we would
| love to collaborate.
| jasonpeacock wrote:
| > and creating humorous photos with celebrities.
|
| how are you handling licensing (use of celebrity or any other
| person's image) and copyright (original image is not from user,
| so is likely copyrighted by someone else)?
| ugh123 wrote:
| Is this not a case of being allowed for personal use and/or
| satire vs publishing the rendered works and _then_ stepping on
| copyright?
| yardenst wrote:
| This is something our customers need to deal with
| JohnFen wrote:
| Depending on jurisdiction, you may find that this isn't
| exactly the case.
| lancesells wrote:
| This is a strange answer when having celebrities on your home
| page and a demo of famous people.
| JohnFen wrote:
| Sigh. This is just the sort of thing that people are afraid that
| AI will be used for. The fear is not entirely unjustified.
| ConnorMooneyhan wrote:
| Here comes Rule 34...
| AdamH12113 wrote:
| It's neat technology, but...
|
| I'm trying to think of uses for this that aren't creepy
| ("manipulate... the pose of the couple... their facial
| expressions..."), dishonest ("photos... with celebrities..."), or
| dystopian ("creating new photo albums from old scanned albums...
| providing personalized content for user acquisition campaigns")
| and failing. If someone comes along and makes a fake photo of
| themself with a celebrity, that's great, but it's also a one-off
| novelty. What kind of non-sleazy recurring use is going to prompt
| people to pay for this?
___________________________________________________________________
(page generated 2023-11-13 23:01 UTC)