Japanese Stable Diffusion
Japanese Stable Diffusion adapts Stable Diffusion to Japanese text and image concepts. rinna first trains a Japanese text encoder, then tunes that encoder and the diffusion model together.
Restricted access
Language adaptation changes the text encoder
The card reports about 100 million images with Japanese captions, including Japanese LAION-5B data. It identifies weaker non-Japanese prompting and difficulty with legible text and complex object arrangements.Japanese Stable Diffusion official model card
Write the scene directly in Japanese
Practical guidance. Use Japanese descriptions for the intended scene and check object placement separately from visual style. For a sign or label that must be exact, inspect every character in the generated result.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Status | Restricted access |
| Launch date | |
| Access | Downloadable |
| Model identifiers | japanese-stable-diffusion |
| License | CreativeML OpenRAIL-M, including its use conditions.Japanese Stable Diffusion official model card |
| Access | Hugging Face requires acceptance of conditions and sharing of account information before file access.Japanese Stable Diffusion official model card |
Related models
- Japanese CLIP ViT-B/16
rinna ยท Japanese image-text encoders
Japanese CLIP ViT-B/16 learns shared image and Japanese-text representations through contrastive image-text training.
Launch date Downloadable