PhotoRoom Releases Open-Source Text-to-Image Model PRX
ylecun · x · 2026-07-14
PhotoRoom states that independence is their choice: while open to collaborating with major AI labs when appropriate, they are also training their own open-source foundation model, PRX. This is a text-to-image model built in Europe and backed by the French government's France 2030 program.
They emphasize that independence doesn't mean building everything in-house, but rather retaining control over critical components: architecture, training design, and the underlying data pipeline. The team also released Part 4 of the PRX series, focusing on the least glamorous yet most crucial component—data—and shared how they built their data pipeline.
More from Multimodal
- Dense instance segmentation via Muse Wonder + SAM 3.1 on Meta's Model API — nikhilaravi · 2026-10-26
- LichtFeld Studio v0.5.4.3 ships community-built SteadySplats, POPSpa and Voxelization — janusch_patas · 2026-10-26
- One prompt to a 2-3 min narrated short: hands-on test of Antigravity video generation — AI_Andrew · 2026-10-26
- Creator makes 10,564-frame animated short entirely with Gemini toolchain — AI_Andrew · 2026-10-26
- FLUX.3 Klein step-distilled to 4-8 steps with CFG baked in — Fit-Structure-6489 · 2026-10-26
- ostris pre-training "Primal VAE": 1-channel latents match modern 8x VAEs via patching — ostrisai · 2026-10-26