What it is
Open 5B text-to-image DiT with strong text rendering, JSON layout control, and 4K-token prompts, plus public weights and fine-tuning support. Generate images with explicit text placement and object layout, long prompts, and readable in-image text using a public 5B text-to-image model with base and SFT checkpoints.
Key results
- Supports prompts up to 4K tokens
- Provides JSON layout control with bbox-based element placement
- Releases public 5B base and SFT checkpoints with LoRA guidance
Shows that a public 5B image model can combine long-prompt support, layout control, and strong text rendering in one stack.