Photoroom

What does it take to manage 8,000 Poshmark listings while working full-time? For Alexandra Mahl, the answer starts with better visuals. Clean, consistent product images are her competitive edge. ⚡️ With Photoroom’s Batch mode, she transforms her listing photos in minutes, ranks on Google Search almost instantly, and has grown her sales by 500%. 💸 Her story reflects what we’re building toward every day: helping businesses of all sizes create high-quality, professional visuals while dramatically speeding up their workflow. Watch Alex share her journey 👇🏾

The $1M deal is a great milestone, but the shift in how marketplaces choose Photoroom is the part that matters. We’re adding a few roles if you’re exploring what’s next. Check out Michelle Belcic 's post 👇

Today, with Roman Frigg, Jon Almazán and the Photoroom ML Team, we kicked off an open-source text-to-image foundation model, PRX, trained from scratch and already available in 🤗 Diffusers. This will be a fully open journey where we’ll share everything we learn along the way—from wins and mistakes to benchmarks and training techniques. Try the model preview and read our first blog posts — links in the comments https://lnkd.in/didCrtDN

A few weeks ago I shared that we're training a text-to-image model from scratch and open-sourcing the entire journey. Today, the first big release is out: PRX weights in 🤗 Diffusers! (Apache 2.0). PRX is a 1.3B-param, flow-matching T2I with a simplified MMDiT backbone, a multilingual text encoder, and separate PRX variants trained with different VAEs and pixel resolutions. Full announcement, demo, and more examples: https://lnkd.in/eFTkhErM We’ll publish follow-up posts soon detailing the design experiments, training choices, benchmarks, and the process behind all this. Huge kudos to the whole ML team at Photoroom, especially Roman Frigg and David Bertoin, who led this effort.

It’s our birthday 🥳🥳🥳 Photoroom turns 6 today 🎂 6 years isn’t a long time in the grand scheme of things, but it’s long enough to see how much can change when you just keep building, keep learning, and keep shipping fast 🥹 We started with a simple goal: make high quality product visuals possible for anyone. No glossy roadmap. Just a problem we thought we could solve better. Since then, the team has grown, the product has grown, and the expectations have definitely grown-- and we’re still experimenting, still breaking things, still shipping features that make us slightly nervous (in a good way). So here’s to six years of progress that wasn’t always linear, solutions that weren’t always scalable, and a heck of a Gen AI wave to ride. And thank you... to our team, our users, and our community for pushing us forward. Cheers to what’s next… the most interesting things are still ahead 🥂 P.S. Can you tell us what is real and what is AI in this picture? 🤪

All Photoroomers are experts in their fields, and within the Data team, that expertise is multiplied by the way we embrace GenAI to accelerate our impact and elevate the way we work 🤖 🧠 Our Head of Data, Juliette D., had the chance to speak with Robin Conquet on DataGen about how we’re reshaping data workflows at Photoroom, and how integrating AI-powered tools into our processes has become part of our team’s DNA, empowering every Photoroomer to explore, understand, and analyze data on their own 🔥 🎧 Listen (or watch with subtitles) wherever you get your podcasts!

We're so excited to see Photoroom featured in Accel's Globalscape 2025 Report! 🥳 As generative AI continues to transform how people create and interact with visual content, we’re proud to be building the next generation of creative tools that empower everyone to sell faster and smarter 🏃♂️➡️ Matt Rouif

Why we revamped our Amplitude Experiments methodology at Photoroom: from t-test to sequential testing. At Photoroom, the data team’s mission is to ensure everyone has access to reliable self-serve analytics. We’ve used Amplitude since day one, and we rely heavily on its Experiments feature; running dozens of A/B tests at the same time, from ML model comparisons to onboarding flows. To make sure everyone could run experiments autonomously, we decided to write internal guidelines. And we learnt a lot. We spent hours reading documentation, discussing with Juliette, and asking questions to Amplitude’s experts. Along the way, we decided to start writing articles to share what we’ve learned. For this first one, we share how we discovered that Amplitude provides two A/B testing methodologies, and why the one we used wasn’t ideal for a PLG company like ours. We completely revamped our approach to experimentation, and we’re excited to share why. https://lnkd.in/efkuPcyv