🏞️ We're training text-to-image models from scratch ⚙️All things Machine Learning & Engineering @ Photoroom
Worldwide 🌎
- Tweets 825
- Following 220
- Followers 7,917
- Likes 1,439
Photoroom retweeted
Happy to see @nvidia include my blog post on how we cut inference cold starts to seconds at @photoroom in the official Dynamo docs 🤩
Pretty cool to have our work referenced there. Thanks, @nvidia!
docs.nvidia.com/dynamo/exter…
3
2
7
357
Fixing one part of an AI image is like playing Jenga.
Our co-founder and CEO @matthieurouif at @HumanXCo Amsterdam on how we fix just the broken part, so brands can trust AI across millions of images 👇
#HumanX #AI
1
2
1,668
Team growth and seller growth move together here. We're hiring 4 product managers in Paris.
Real B2B experience. Real diffusion model and LLM shipping experience. A creative tools or design background with real craft sensibility.
4 roles, one thread. Links below to apply.
2
2
513
Photoroom retweeted
Astra is quite impressive with iOS apps 🤩
I asked it to build me an app that takes pictures of Pokémon cards, with perfect perspective correction.
It used the @photoroom API to remove the background and then implemented dewarping using the optical characteristics of the iPhone camera.
Took me less than an hour vs a few days if I'd had to go through the dewarping math myself.
I’ve spent the afternoon going to Pokémon card shops in Tokyo and asking ChatGPT to find me the cards that could have a good resell price in Europe.
Now I wish I could stay for a month and build an actual app that does this for @RevenueCat Shipaton 🤪
14
10
181
42,019
One AI model can create the visual. But who checks if it got it right? Visual Agents brings multiple agents into one tailored workflow to analyze, create, score and fix every output before it reaches your buyers.
More on how it works in the comments.
1
2
486
Learn more about Visual Agents: photoroom.com/inside-photoro…
243
Reasoning made LLMs better. We think image editing is next. That's what we're building at Photoroom: the fidelity layer. Before generating, the model figures out exactly what has to be preserved. After generating, it checks its own output against that, and corrects itself, instead of trusting a single forward pass. Same bet as every reasoning model: don't trust one pass. Give the system room to think, verify, and fix its own work.
2
1
3
3,216
The research behind the fidelity layer, in full → photoroom.com/blog/introduci…
1
300
Watch our ML team discuss why product fidelity matters: piped.video/watch?v=pYev4cSE…
168