v0.3.51

Comfy-Org/ComfyUIv0.3.51Aug 20, 2025by github-actions[bot]

AI Summary

This release marks the implementation of the Qwen Image Edit model and adds support for various new API nodes including Minimax, GPT-5, and Vidu.

Key Highlights

Implementation of Qwen Image Edit model
Support for Wan 2.2 camera model
API nodes for Minimax HailuoVideo and GPT-5 series
API nodes for Vidu Video

New Features

Qwen Image Edit model
Wan 2.2 camera model
API nodes for Minimax, GPT-5, and Vidu
FluxKontextMultiReferenceLatentMethod node

Full Release Notes

## What's Changed
* Put back frontend version. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9317
* Reduce portable size again. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9323
* SDPA backend priority by @contentis in https://github.com/comfyanonymous/ComfyUI/pull/9299
* Make last PR not crash comfy on old pytorch. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9324
* Fix XPU iGPU regressions by @simonlui in https://github.com/comfyanonymous/ComfyUI/pull/9322
* Added context window support to core sampling code by @Kosinkadink in https://github.com/comfyanonymous/ComfyUI/pull/9238
* av is an essential dependency. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9341
* Make custom node testing checkbox optional in issue templates by @webfiltered in https://github.com/comfyanonymous/ComfyUI/pull/9342
* Update CODEOWNERS by @yoland68 in https://github.com/comfyanonymous/ComfyUI/pull/9343
* Update default parameters for Moonvalley video nodes by @guill in https://github.com/comfyanonymous/ComfyUI/pull/9290
* convert WAN nodes to V3 schema by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9201
* Make SLG nodes work on Qwen Image model. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9345
* Avoid torch compile graphbreak for older pytorch versions by @StrongerXi in https://github.com/comfyanonymous/ComfyUI/pull/9344
* Fix last commit not working on older pytorch. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9346
* Add warning when using old pytorch. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9347
* fix(OpenAIGPTImage1): set correct MIME type for multipart uploads by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9348
* Add FluxKontextMultiReferenceLatentMethod node. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9356
* Implement wan2.2 camera model. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9357
* Update frontend to v1.25.8 by @christian-byrne in https://github.com/comfyanonymous/ComfyUI/pull/9361
* record audio node by @jtydhr88 in https://github.com/comfyanonymous/ComfyUI/pull/8716
* Qwen image model refactor. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9375
* WIP Qwen edit model: The diffusion model part. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9383
* Update template to 0.1.60 by @comfyui-wiki in https://github.com/comfyanonymous/ComfyUI/pull/9377
* Make context step index detection much more robust by @Kosinkadink in https://github.com/comfyanonymous/ComfyUI/pull/9392
* Bump frontend to 1.25.9 by @christian-byrne in https://github.com/comfyanonymous/ComfyUI/pull/9394
* fix(WAN-nodes): invalid nodeid for WanTrackToVideo by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9396
* P2 of qwen edit model. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9412
* api_nodes: add kling-v2-1 and v2-1-master by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9257
* api_nodes: add MinimaxHailuoVideoNode node by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9262
* api_nodes: add GPT-5 series models by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9325
* api_nodes: release version of google's models by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9304
* feat(api-nodes): add Vidu Video nodes by @bigcat88 in https://github.com/comfyanonymous/ComfyUI/pull/9368
* Change the TextEncodeQwenImageEdit node to use logic closer to refererence code. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9432
* Rope fix for qwen vl. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9435
* Qwen rotary embeddings should now match reference code. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9437
* Disable prompt weights for qwen. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9438
* Fix bf16 precision issue with qwen image embeddings. by @comfyanonymous in https://github.com/comfyanonymous/ComfyUI/pull/9441
* Bump template to 0.1.62 by @comfyui-wiki in https://github.com/comfyanonymous/ComfyUI/pull/9419

## New Contributors
* @StrongerXi made their first contribution in https://github.com/comfyanonymous/ComfyUI/pull/9344

**Full Changelog**: https://github.com/comfyanonymous/ComfyUI/compare/v0.3.50...v0.3.51