AuroraIMG2-50M-Preview

Use it to generate 256×256 images from English text prompts or to study and experiment with small-scale text-to-image training.

Availability

Preview release on Hugging Face

Pricing model

Not established

Deployment

Load the model with Diffusers

Privacy

Not established

Licence

Apache 2.0 for model weights and code; check dataset terms for commercial use

• The model generates fixed-resolution 256×256 RGB images from text prompts. ^[1]
• The model is released as a preview on Hugging Face with a Diffusers quick-start. ^[1]
• The released model includes a 50.4M-parameter denoiser and a 12.6M-parameter text encoder. ^[1]

  AuroraIMG2-50M-Preview is a 256×256 text-to-image model built around a **50.4M-parameter flow-matching diffusion transformer** and a **12.6M-parameter text encoder**.

• The model supports 50 sampling steps by default and 20 steps for faster generation. ^[1]

  - Supra2-IMG and Surjo-Image-Preview numbers are as reported on the Surjo-Image-Preview model card (256px, 50 steps, CFG 3.0, seed 42, one image per prompt).

• The model is English-only and supports prompts of up to 62 caption tokens. ^[1]
• The model card says it is not for generating realistic images of real people or events, or for settings where accuracy and fairness matter. ^[1]

  - **Not for** generating realistic images of real people or events, or for settings where accuracy and fairness matter.