Zum Inhalt springen
Node packGesehenExtern
Quelle öffnen

sweet-tea-nodes

Sweet Tea custom node pack for ComfyUI registry

Betreuer@tea-time-labsD · adoptionBewertung

Paketprofil

Status
Gesehen
Verleger
@tea-time-labs
Quelle
GitHub
Sterne
0
Verknüpftes Pipes
2

README

Sweet Tea Nodes

ComfyUI node pack focused on high-control video generation and high-fidelity tiled upscaling.

Node Highlights

  • Ultimate SD Upscale (Pose Tiled ControlNet) (UltimateSDUpscalePoseTiled)
    • Major tiled-upscaling upgrade with lockable padding, ControlNet tile alignment fixes, normalized overlap blending, and seam-fix passes designed to remove bleed and ghost offsets.
  • Wan VACE -> Video (caps inpaint) (WanVaceToVideoCapsInpaint)
    • Temporal cap-stitching node that treats start/end caps as known regions and forces middle-region generation with explicit inpaint mask semantics.
    • Built for precise clip stitching and transition reconstruction workflows.
  • Wan VACE -> Video (hybrid subject swap) (WanVaceToVideoHybridSubjectSwap)
    • Depth-first hybrid subject-swap node for Wan VACE workflows.
    • Combines structural control with a character reference and optional source-RGB luma/detail/background contexts for controllable illustrated/live-action blending.

Source Provenance

This repository packages node code synced from the sweettea R2 bucket paths:

  • custom_nodes/usdu_pose_tiled/*
  • custom_nodes/wan_vace_caps_inpaint/*
  • custom_nodes/wan_vace_hybrid_subject/*

Install (ComfyUI)

  1. Open ComfyUI Manager.
  2. Search for Sweet Tea Nodes in the registry (after publish), then install.
  3. Restart ComfyUI.

Manual install from Git:

cd ComfyUI/custom_nodes
git clone https://github.com/tea-time-labs/sweet-tea-nodes.git

Local Validation Commands

Run from repo root:

python3 -m py_compile __init__.py wan_vace_caps_inpaint/*.py wan_vace_hybrid_subject/*.py usdu_pose_tiled/*.py
comfy --skip-prompt --here node validate
comfy --skip-prompt --here node pack

Release Automation

This repo includes two GitHub workflows:

  • Validate Node Pack: compile checks + comfy node validate on push and PR.
  • Release and Publish: automatic semantic version bump from commit history, git tag creation, GitHub release creation, and Comfy Registry publish.

Autoversion rules in main pushes:

  • BREAKING CHANGE or type!: -> major bump
  • feat: -> minor bump
  • everything else -> patch bump

Required secret for registry publish:

  • REGISTRY_ACCESS_TOKEN

Notes

  • WanVaceToVideoCapsInpaint supports variable lengths and cap sizes.

Knoten in diesem Paket

7 Knoten (Plural)
  • SweetTeaExecutionReceipt
  • SweetTeaPreviewImage
  • SweetTeaPreviewVideo
  • SweetTeaPreviewVideoFromImages
  • UltimateSDUpscalePoseTiled
  • WanVaceToVideoCapsInpaint
  • WanVaceToVideoHybridSubjectSwap

Verwandte Ressourcen

40 Verbindungen
Verwendet vonVidu Q3 image to videoVerwendet vonWan 2.2 first–last frame videoVerwendet vonKrea 2 Turbo text to imageVerwendet vonQwen Image 2512 text to imageVerwendet vonImage upscale and refinementVerwendet vonFLUX virtual try-onVerwendet vonHiDream I1 Fast image generatorVerwendet vonWan 2.1 video transformationVerwendet vonRemove backgrounds with BiRefNetVerwendet vonQwen Image 2512 · 360 PanoramaVerwendet vonGoogle Nano Banana 2Verwendet vonMagazine packaging presentationVerwendet vonFLUX.1 Krea image generationVerwendet vonAdd music to a videoVerwendet vonZ-Image upscale and refineVerwendet vonCreate a HeyGen talking-head videoVerwendet vonKandinsky 5 text to videoVerwendet vonWan Animate 2: Motion transferVerwendet vonExploded view round tripVerwendet vonZ-Image Turbo text to imageVerwendet vonLens Turbo text to imageVerwendet vonLuma video with camera motionVerwendet vonAnimate an image with Vidu Q2Verwendet vonVidu Q1 text to videoVerwendet vonSketch and collage to imageVerwendet vonIdeogram 4 image generatorVerwendet vonSeedVR2 image upscaleVerwendet vonDALL·E 2 image generationVerwendet vonFirst and last frame to video · LTX-2.5Verwendet vonHunyuanVideo text to videoVerwendet vonGemini reference images to videoVerwendet vonProduct placement photoVerwendet vonZ-Image Turbo with PiD 4× upscaleVerwendet vonCombined character sheetVerwendet vonMagazine coverVerwendet vonSwap product into a reference adVerwendet vonQwen Image EditVerwendet vonDieline to box mockupVerwendet vonPhoto to geometry mapsVerwendet vonFlux Kontext image editor

Quellen

3 Quellen
GitHubQuelle
github.com/tea-time-labs/sweet-tea-nodes

Verifiziert 4. Nov.

ManualQuelle

Quellenauszüge

1 Auszug

Source context: Repo tea-time-labs/sweet-tea-nodes