PrunaAI/P-H3-Image
AI model · https://huggingface.co/PrunaAI/P-H3-Image
First seen by Oriel on .
AI summary
This model card introduces P-H3-Image, a MiniMax-H3-based pipeline for 8-megapixel text-to-image generation and multi-image editing in eight sampling steps without classifier-free guidance.
- The bf16 package includes a 33.1B-parameter transformer, 4B text encoder, VAE, and Qwen3.5-9B-based prompt upsampler, totaling 100 GB.
- It accepts up to eight images; Picture 1 is edited while the others serve as named references for composing changes.
- At 8 MP on one NVIDIA B200, generation takes about 9 seconds and 83 GB; upsampling adds 6–11 seconds and raises memory to 101 GB.
- Auto CPU offloading reduces peak usage to about 76 GB, while separate int8 and w4a8 ComfyUI files target 21 GB VRAM.
- The license excludes the European Union, United Kingdom, Republic of Korea, and United States without separate MiniMax licensing; outputs also struggle with hands, small people, and dense text.
#text-to-image #image editing #diffusers #model optimization
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 369, latest observation ; +264 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 2,104, latest observation ; +1,817 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | PrunaAI/P-H3-Image | https://huggingface.co/PrunaAI/P-H3-Image Read discussion |