Hugging Face· PlatformResearch paper

PRX Part 3 — Training a Text-to-Image Model in 24h!

Photoroom describes a 24-hour training speedrun for a text-to-image model using a combination of techniques such as x-prediction in pixel space, perceptual losses, token routing with TREAD, and representation alignment with REPA, and open-sources the code.

From the source

we’re doing a 24-hour speedrun: 32 H200 ~ $1500 total compute budget

huggingface.co

Announced

03 Mar 2026 16:50 UTC

Detected

03 Aug 2026 21:12 UTC (+153d after)