{"full_name":"huggingface/diffusers","owner":"huggingface","name":"diffusers","shard":"1","url":"https://github.com/huggingface/diffusers","description":"🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch.","homepage":"https://huggingface.co/docs/diffusers","topics":["deep-learning","diffusion","image-generation","pytorch","score-based-generative-modeling","image2image","text2image","stable-diffusion","stable-diffusion-diffusers","latent-diffusion-models","flux","image2video","text2video","video2video","qwen-image"],"tag_pages":["deep-learning","diffusion","flux","image-generation","pytorch","stable-diffusion","text2image"],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/25720743?v=4","created_at":"2022-05-30T16:04:02Z","pushed_at":"2026-11-15T04:24:53Z","metrics":{"stars":34917,"forks":7480,"watchers":219,"issues":5751,"prs":8491,"releases":97,"issues_open":942,"issues_closed":4809,"prs_open":658,"prs_closed":1545,"prs_merged":6288,"discussions":677},"growth":{"day":{"stars":-1,"forks":1,"base_day":"2026-11-15"},"week":{"stars":27,"forks":10,"base_day":"2026-11-09"}},"history":{"dates":["2026-11-02","2026-11-03","2026-11-04","2026-11-09","2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[34844,34851,34856,34890,34894,34908,34914,34914,34918,34918,34917],"forks":[7437,7451,7455,7470,7475,7478,7478,7477,7479,7479,7480],"watchers":[218,219,219,220,220,220,220,220,219,219,219],"issues_open":[null,null,null,null,null,920,928,930,936,940,942],"prs_open":[null,null,null,null,null,648,652,649,651,653,658]},"boards":{},"same_owner":[{"full_name":"huggingface/transformers","name":"transformers","shard":"1","boards":{"total":54}},{"full_name":"huggingface/OpenEnv","name":"OpenEnv","shard":"1","boards":{}},{"full_name":"huggingface/accelerate","name":"accelerate","shard":"1","boards":{}},{"full_name":"huggingface/agents-course","name":"agents-course","shard":"1","boards":{}},{"full_name":"huggingface/alignment-handbook","name":"alignment-handbook","shard":"1","boards":{}},{"full_name":"huggingface/ml-intern","name":"ml-intern","shard":"1","boards":{}},{"full_name":"huggingface/parler-tts","name":"parler-tts","shard":"1","boards":{}},{"full_name":"huggingface/pytorch-image-models","name":"pytorch-image-models","shard":"1","boards":{}},{"full_name":"huggingface/sentence-transformers","name":"sentence-transformers","shard":"1","boards":{}},{"full_name":"huggingface/skills","name":"skills","shard":"1","boards":{}},{"full_name":"huggingface/smolagents","name":"smolagents","shard":"1","boards":{}},{"full_name":"huggingface/speech-to-speech","name":"speech-to-speech","shard":"1","boards":{}},{"full_name":"huggingface/transformers.js","name":"transformers.js","shard":"1","boards":{}},{"full_name":"huggingface/trl","name":"trl","shard":"1","boards":{}}],"issues":[{"number":9108,"title":"can't resume training with quantized base model","comments":45,"created_at":"2024-08-07T04:25:34Z"},{"number":5808,"title":"Inpainting produces results that are uneven with input image","comments":43,"created_at":"2023-11-15T16:33:02Z"},{"number":13401,"title":"Help us profile important pipelines and improve if needed","comments":42,"created_at":"2026-04-03T17:18:13Z"},{"number":9924,"title":"Can we get more schedulers for flow based models such as SD3, SD3.5, and flux","comments":40,"created_at":"2024-11-14T00:07:56Z"},{"number":9180,"title":"Pipeline has no attribute '_execution_device'","comments":38,"created_at":"2024-08-14T14:43:15Z"},{"number":9279,"title":"Support dynamic LoRA loading with `torch.compile` model","comments":37,"created_at":"2024-08-26T15:10:04Z"},{"number":7219,"title":"Add SUPIR Upscaler","comments":36,"created_at":"2024-03-05T09:32:38Z"},{"number":5894,"title":"Error RuntimeError: Invalid buffer size ","comments":36,"created_at":"2023-11-22T11:36:01Z"}],"discussions":[{"number":15255,"title":"loraspectra: Fast Rust CLI to audit, compress, convert (PEFT/Kohya/Diffusers), and orthogonally merge Diffusion LoRA/DoRA .safetensors","comments":0,"category":"Show and tell","created_at":"2026-11-12T02:30:46Z"},{"number":14951,"title":"What is the easiest Diffusers pipeline to start with for a beginner?","comments":1,"category":"Q&A","created_at":"2026-10-05T17:56:05Z"},{"number":14893,"title":"Real-time serving on diffusers with LTX-2.5 (22B video DiT, single GPU) — notes and two negative results","comments":0,"category":"Show and tell","created_at":"2026-09-28T09:23:08Z"},{"number":14857,"title":"One obsession text encoder","comments":4,"category":"Q&A","created_at":"2026-09-23T15:22:38Z"},{"number":14823,"title":"A four-layer evidence pattern for apps that persist Diffusers outputs","comments":0,"category":"Show and tell","created_at":"2026-09-20T20:09:08Z"},{"number":14813,"title":"How to generate white background images instead of colorful?","comments":3,"category":"Q&A","created_at":"2026-09-19T20:28:49Z"},{"number":14251,"title":"[Modular Diffusers] Repo ID to custom path mapping","comments":0,"category":"General","created_at":"2026-07-20T16:07:45Z"},{"number":14173,"title":"🎬 [Showcase] AI-Generated Anime Short - 'Light in the Ruins'","comments":0,"category":"Show and tell","created_at":"2026-07-12T04:51:20Z"}],"latest_release":{"name":"Diffusers 0.43.0: MiniMax-H3-Image, Pika H3 Flex, P-H3-Image, Kandinsky S2V, and PAN 2","tag":"v0.43.0","published_at":"2026-11-12T05:22:22Z","body":"> [!TIP]\r\n> This release adds five modular model integrations for image generation, editing, video with audio, speech-driven animation, and world simulation. It also improves attention backends, tensor parallelism, LoRA handling, and checkpoint saving.\r\n\r\n## MiniMax-H3-Image\r\n\r\nMiniMax-H3-Image supports text-to-image generation and editing with one or more reference images. The integration includes a dedicated image transformer and autoencoder, with a fixed six-step sampling schedule and distilled guidance. A separate prompt-enhancement workflow expands short requests and suggests an aspect ratio.\r\n\r\nLoad it through `ModularPipeline.from_pretrained(...)`, then call `load_components(...)`. The [MiniMax-H3-Image documentation](https://huggingface.co/docs/diffusers/main/en/api/pipelines/minimax_h3_image) covers reference-image prompts, output sizes, prompt enhancement, and component offloading.\r\n\r\n## Pika H3 Flex\r\n\r\nPika H3 Flex builds on MiniMax-H3 to generate video and synchronized stereo audio in six steps. It supports text prompts, first and/or last frames, and multimodal references through separate modular workflows. The integration includes LoRA adapter loading and a guide to sharing already-loaded MiniMax-H3 components between the two pipelines.\r\n\r\nSee the [Pika H3 Flex documentation](https://huggingface.co/docs/diffusers/main/en/api/pipelines/pika_h3_flex) for workflow selection, adapter loading, and component reuse.\r\n\r\n## P-H3-Image\r\n\r\nP-H3-Image from Pruna AI brings text-to-image generation and instruction-based editing with up to eight input images. Its modular pipeline includes a prompt upsampler that turns requests into detailed generation or editing prompts, an image autoencoder, and a text connector. The checkpoint uses eight sampling steps with distilled guidance.\r\n\r\nSee the [P-H3-Image documentation](https://huggingface.co/docs/diffusers/main/en/api/pipelines/p_h3_image) for image sizing, reference-image editing, prompt upsampling, and offloading. Thanks to @nifleisch for the integration in [#15171](https://github.com/huggingface/diffusers/pull/15171).\r\n\r\n## Kandinsky S2V 1.0\r\n\r\nKandinsky S2V 1.0 animates a person from a reference image and a speech recording, including full-body movement aligned with the speech. It generates video in five-second chunks, carrying FramePack memory between chunks. The supplied speech provides the soundtrack; the model generates the video.\r\n\r\nThe modular integration supports Lite and Pro checkpoints, including distilled variants that use `PiflowScheduler` with a default of 16 steps. See the [Kandinsky S2V documentation](https://huggingface.co/docs/diffusers/main/en/api/pipelines/kandinsky_s2v) for audio preparation, checkpoint settings, and memory requirements. Thanks to @nvsm for the integration in #15164.\r\n\r\n## PAN 2\r\n\r\nPAN 2 simulates scenes from prompts describing actions and camera movements over time. This release integrates its 29B generative simulator through text-to-video and image-to-video modular workflows. Context parallelism and FirstBlockCache, MagCache, and TaylorSeer caching are supported. Prompt and video safety checks are enabled by default and require the separate `pan2-guardrail` package.\r\n\r\nSee the [PAN 2 documentation](https://huggingface.co/docs/diffusers/main/en/api/pipelines/pan2) for loading, guardrails, and distributed inference. The integration covers the simulator; the separate reasoning model that writes its prompts is linked from the model documentation. Thanks to @wu-qing-157 for the integration in #15202.\r\n\r\n### Other highlights\r\n\r\n- **MiniMax-H3 tensor parallelism:** shard the transformer across devices, including its large modulation projections. Replicated parameters also retain their intended dtype during tensor-parallel loading. See: [#14609](https://github.com/huggingface/diffusers/pull/14609), [#15147](https://github.com/huggingface/diffusers/pull/15147).\r\n- **AMD attention support:** the new `aiter_sage_hub` backend brings a Hub-hosted SageAt","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}