Skip to content

UPSTREAM_PINS.md

Latest commit

 

History

History
44 lines (33 loc) · 3.28 KB

UPSTREAM_PINS.md

File metadata and controls

44 lines (33 loc) · 3.28 KB
Raw

Upstream pins

This fork follows upstream projects deliberately. Updates are reviewed and tested; they are not pulled automatically into a working branch.

Application base

Project Upstream Branch Pinned commit Commit date License Purpose
OpenScene https://github.com/Theorvane/openscene dev 86e4be3ae2eeed51ba48abaa9df2a8bce46b6715 2026-08-29 MIT Desktop editor, shared timeline, provider seams, job management and local export

The local clone keeps OpenScene as upstream; the writable origin is the personal fork at https://github.com/3ongtam-coder/openscene.git.

Runtime baseline

Component Baseline Status
Windows Current development machine Active
Node.js v24.14.1 Root and mobile typechecks pass
npm 11.11.0 Root and mobile dependencies installed with npm ci
FFmpeg/FFprobe 9.0.1-full_build-www.gyan.dev Installed system-wide with winget; GPL/version 3 enabled; do not bundle into a closed distribution without a separate license decision
GPU NVIDIA GeForce GTX 1650, 4 GB VRAM, compute capability 7.5 Insufficient for the planned Wan 2.2 14B local workflow; use a remote worker or cloud lane

Model and workflow pins

OpenScene now contains a generic ComfyUI API-workflow adapter, but does not vendor or install the worker, nodes, workflows, or weights. Add an exact commit/tag, model-weight checksum and weight license to a deployment record before enabling a workflow in a release.

Project Planned role Current decision
ComfyUI Remote/local workflow worker API adapter implemented; deployment remains a separate process/service and must pin a stable release, not master
ComfyUI-WanVideoWrapper Wan workflow nodes Supported through user-supplied API workflow; pin with every workflow bundle and run golden tests
Wan 2.2 Motion control and character retarget Move/Mix seam implemented; use a remote GPU on the current machine, and record code/weight licenses separately
whisper.cpp Local subtitle transcription Managed Windows x64 release b4938 / version 1.9.3 (MIT), release archive SHA-256 c2a4b60edb11f7e11a9191ffb50929535527d4d91c9903dbe3e554583bbbc63d; the adapter still follows reviewed commit 52a939a2a762224e255d366c1182b2af4dd1a032
ggml-small.bin Multilingual local subtitle model ggerganov/whisper.cpp commit 5359861c739e955e79d9a303bcbc70fb988958b1 (MIT metadata), SHA-256 1be3a9b2063867b937e64e2ec7483364a79917e157fa98c5d94b5c1fffea987b; installed only under ignored .local-runtimes/
VieNeu-TTS v3 Turbo Local Vietnamese narration Separate checkout pnnbao97/VieNeu-TTS commit 6c12b81a369ac61ef906f1c936b5c983f2603660 (Apache-2.0 code); OpenScene synchronizes its .venv and launches its loopback demo, while model/voice terms remain non-commercial and separately governed

Update procedure

  1. Fetch upstream without merging.
  2. Review release notes, migrations and license changes.
  3. Create an issue-scoped upgrade branch from upstream/dev.
  4. Change one upstream layer per pull request.
  5. Run root typecheck, root tests, root build and mobile typecheck.
  6. Run media and workflow golden fixtures after those fixtures exist.
  7. Update this file only after the upgrade is accepted.