Skip to content

ComfyUI

v0.22.0 Breaking

This release includes 2 breaking changes for platform teams planning a safe upgrade.

Published 14d Model Serving & MLOps
✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

ai comfy comfyui python pytorch stable-diffusion

ReleasePort's take

Moderate signal
editorial:auto 14d

ComfyUI v0.22.0 adds MoGe, HiDream-O1 area conditioning, Stable Audio 3 support, and fixes critical crashes in Hunyuan3D batch size and Qwen3.5 multi-image text generation. /api/upload/mask endpoint is deprecated in favor of /api/upload/image.

Why it matters: Developers must migrate from /api/upload/mask to /api/upload/image. Production users running Hunyuan3D or Qwen3.5 multi-image workflows should upgrade immediately for stability. LTX2.3 VRAM reduction with guide_mask benefits inference efficiency.

Summary

AI summary

Broad release touches openapi, feat, CORE-198, and chore.

Changes in this release

Security High

Create SECURITY.md

Create SECURITY.md

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Feature Medium

Support MoGe (CORE-168)

Support MoGe (CORE-168)

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Feature Medium

Add downscaled IC-LoRA support to LTXVAddGuide (CORE-102)

Add downscaled IC-LoRA support to LTXVAddGuide (CORE-102)

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Feature Medium

Add optional attention_mask input to LTXVAddGuide (CORE-220)

Add optional attention_mask input to LTXVAddGuide (CORE-220)

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Feature Medium

Initial work to make downscale_ratio_temporal work

Initial work to make downscale_ratio_temporal work

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Feature Medium

feat(openapi): add optional description field to workspace API key schemas

feat(openapi): add optional description field to workspace API key schemas

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Feature Medium

Adding new StringFormat node

Adding new StringFormat node

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Feature Medium

Support Stable Audio 3 model

Support Stable Audio 3 model

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Feature Medium

Update workflow templates to v0.9.79

Update workflow templates to v0.9.79

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Feature Medium

HiDream-O1: support area conditioning

HiDream-O1: support area conditioning

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Feature Medium

Use temporal downscale to make empty audio latent nodes more reusable

Use temporal downscale to make empty audio latent nodes more reusable

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Feature Medium

Make audio processing nodes handle None inputs

Make audio processing nodes handle None inputs

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Feature Medium

Preserve noise_scale/shift across chained model_sampling patches

Preserve noise_scale/shift across chained model_sampling patches

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Feature Medium

Implement/fix dynamic clip saving

Implement/fix dynamic clip saving

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Feature Low

Add careers link to README and startup log

Add careers link to README and startup log

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Feature Low

[Partner Nodes] add ByteDance Seed LLM node

[Partner Nodes] add ByteDance Seed LLM node

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Performance Medium

Reduce LTX2.3 peak VRAM when guide_mask is in use (CORE-166)

Reduce LTX2.3 peak VRAM when guide_mask is in use (CORE-166)

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Performance Medium

Reduce min for Batch Image/Mask/Latent nodes from 2 to 1

Reduce min for Batch Image/Mask/Latent nodes from 2 to 1

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Performance Medium

Add spacial downscale ratio to empty ltxv latent

Add spacial downscale ratio to empty ltxv latent

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Deprecation Medium

docs(openapi): deprecate /api/upload/mask in favor of /api/upload/image

docs(openapi): deprecate /api/upload/mask in favor of /api/upload/image

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Bugfix Medium

Fix Qwen3.5 text generation with multi-image prompts

Fix Qwen3.5 text generation with multi-image prompts

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Bugfix Medium

[Partner Nodes] fix Opus 4.7 sending deprecated temperature parameter

[Partner Nodes] fix Opus 4.7 sending deprecated temperature parameter

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Bugfix Medium

Fix BiRefNet issue

Fix BiRefNet issue

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: high

 —
Bugfix Medium

fix: Hunyuan3D 2.1 batch size crashes in attention and forward pass

fix: Hunyuan3D 2.1 batch size crashes in attention and forward pass

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Bugfix Medium

Allow negative `batch_index` on ImageFromBatch and LatentFromBatch (CORE-195)

Allow negative `batch_index` on ImageFromBatch and LatentFromBatch (CORE-195)

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Bugfix Medium

Fix typo in ops.py

Fix typo in ops.py

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Bugfix Medium

Revert inclusion of workflow_id in execution WebSocket messages (CORE-198)

Revert inclusion of workflow_id in execution WebSocket messages (CORE-198)

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Bugfix Low

Fix FeatherMask negative zero indexing for right/bottom feathering

Fix FeatherMask negative zero indexing for right/bottom feathering

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Refactor Medium

Various QoL updates of nodes display names, descriptions and categories (CORE-190, CORE-191)

Various QoL updates of nodes display names, descriptions and categories (CORE-190, CORE-191)

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Refactor Low

Remove annoying message

Remove annoying message

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Refactor Low

Generalize frontend version warning to all comfy* requirements.txt entries

Generalize frontend version warning to all comfy* requirements.txt entries

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Other Medium

docs(openapi): remove top-level width/height fields on Asset schema

docs(openapi): remove top-level width/height fields on Asset schema

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Other Medium

docs(openapi): tighten workspace API key description field (BE-1004)

docs(openapi): tighten workspace API key description field (BE-1004)

Source: granite4.1:8b-q6_K@2026-05-20

Confidence: low

 —
Other Low

Add test that each model has unique identifiers (CORE-134)

Add test that each model has unique identifiers (CORE-134)

Source: granite4.1:30b@2026-05-20-audit

Confidence: low

 —
Full changelog

What's Changed

  • Add test that each model has unique identifiers (CORE-134) by @Talmaj in https://github.com/Comfy-Org/ComfyUI/pull/13654
  • Make audio processing nodes handle None -inputs by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/13879
  • Update workflow templates to v0.9.77 by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/13895
  • Add careers link to README and startup log by @robinjhuang in https://github.com/Comfy-Org/ComfyUI/pull/13897
  • Remove annoying message. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13899
  • Include workflow_id in all execution WebSocket messages (CORE-198) by @christian-byrne in https://github.com/Comfy-Org/ComfyUI/pull/13684
  • Revert "Include workflow_id in all execution WebSocket messages (CORE-198)" by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13901
  • Create SECURITY.md. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13902
  • Generalize frontend version warning to all comfy* requirements.txt entries by @christian-byrne in https://github.com/Comfy-Org/ComfyUI/pull/13875
  • Preserve noise_scale/shift across chained model_sampling patches by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/13892
  • feat: Support MoGe (CORE-168) by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/13878
  • Allow negative batch_index on ImageFromBatch and LatentFromBatch (CORE-195) by @drozbay in https://github.com/Comfy-Org/ComfyUI/pull/13857
  • Reduce LTX2.3 peak VRAM when guide_mask is in use (CORE-166) by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/13735
  • Remove useless option and clarify what lowvram does. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13922
  • feat: Add downscaled IC-LoRA support to LTXVAddGuide (CORE-102) by @drozbay in https://github.com/Comfy-Org/ComfyUI/pull/13896
  • Enhance README with application and cloud links by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13936
  • Log which quant ops are enabled/emulated. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13946
  • chore: correct OOM memory summary format by @Apophis3158 in https://github.com/Comfy-Org/ComfyUI/pull/13950
  • Fix Qwen3.5 text generation with multi-image prompts by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/13943
  • HiDream-O1: support area conditioning by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/13944
  • [Partner Nodes] fix Opus 4.7 sending deprecated temperature parameter by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/13955
  • fix(FeatherMask): correct negative zero indexing for right/bottom feathering by @alvinttang in https://github.com/Comfy-Org/ComfyUI/pull/12881
  • implement/fix dynamic clip saving by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/13959
  • [Partner Nodes] add ByteDance Seed LLM node by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/13919
  • Fix BiRefNet issue by @yousef-rafat in https://github.com/Comfy-Org/ComfyUI/pull/13966
  • feat: Add optional attention_mask input to LTXVAddGuide (CORE-220) by @drozbay in https://github.com/Comfy-Org/ComfyUI/pull/13965
  • Reduce min for Batch Image/Mask/Latent nodes from 2 to 1 by @Kosinkadink in https://github.com/Comfy-Org/ComfyUI/pull/13721
  • Initial work to make downscale_ratio_temporal work. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13972
  • chore: Various QoL updates of nodes display names, descriptions and categories (CORE-190, CORE-191) by @alexisrolland in https://github.com/Comfy-Org/ComfyUI/pull/13830
  • Use temporal downscale to make empty audio latent nodes more reusable. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13975
  • docs(openapi): deprecate /api/upload/mask in favor of /api/upload/image by @mattmillerai in https://github.com/Comfy-Org/ComfyUI/pull/13968
  • docs(openapi): remove top-level width/height fields on Asset schema by @mattmillerai in https://github.com/Comfy-Org/ComfyUI/pull/13973
  • Fix typo in ops.py by @NAM-MAN in https://github.com/Comfy-Org/ComfyUI/pull/11925
  • feat(openapi): add optional description field to workspace API key schemas by @mattmillerai in https://github.com/Comfy-Org/ComfyUI/pull/13993
  • docs(openapi): tighten workspace API key description field (BE-1004) by @mattmillerai in https://github.com/Comfy-Org/ComfyUI/pull/13996
  • Adding new StringFormat node by @Pauan in https://github.com/Comfy-Org/ComfyUI/pull/13997
  • Add spacial downscale ratio to empty ltxv latent. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/13999
  • fix: Hunyuan3D 2.1 batch size crashes in attention and forward pass by @Kivylius in https://github.com/Comfy-Org/ComfyUI/pull/13699
  • Support Stable Audio 3 model. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/14010
  • Update workflow templates to v0.9.79 by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/14011

New Contributors

  • @Apophis3158 made their first contribution in https://github.com/Comfy-Org/ComfyUI/pull/13950
  • @NAM-MAN made their first contribution in https://github.com/Comfy-Org/ComfyUI/pull/11925
  • @Pauan made their first contribution in https://github.com/Comfy-Org/ComfyUI/pull/13997
  • @Kivylius made their first contribution in https://github.com/Comfy-Org/ComfyUI/pull/13699

Full Changelog: https://github.com/Comfy-Org/ComfyUI/compare/v0.21.1...v0.22.0

Breaking Changes

  • Deprecate /api/upload/mask in favor of /api/upload/image (openapi)
  • Remove top-level width/height fields from Asset schema (openapi)
View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track ComfyUI

Get notified when new releases ship.

Sign up free

About ComfyUI

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.

All releases →

Related context

Earlier breaking changes

  • v0.21.1 Revert breaking changes from previous release
  • v0.21.0 Make ColorTransfer node ref_image parameter mandatory
  • v0.21.0 Remove Moonvalley API nodes from codebase
  • v0.21.0 Remove IPEX Intel Extension for Pytorch support

Beta — feedback welcome: [email protected]