SceneMachine β€” Validation Gallery

Validated mp4 outputs from the SceneMachine video stack. Each clip was generated through the production SceneMachine ComfyUIProvider chain (no mocks) and corresponds to a recorded bloodwork test in SceneMachine/operations-log/bloodwork/.

Why this exists

Two reasons:

  1. Proof-of-quality for the stack integrations done in PRs #36-#42. Future agents and contributors can SEE what each Wan / LTX path produces at known-good settings.
  2. Visual regression baseline. If a future PR touches a workflow builder, generating the same prompt + seed should produce the same shot. Drift can be diff'd against these baselines.

Contents

See MANIFEST.json for full per-entry metadata (patient, prompt, settings, wallclock, doctor).

ID Patient Wallclock Size Notes
eb7a6bae wan22-t2v-fp8 57.5 s 791 KB First T2V validation
8ba2fd33 wan22-animate-bf16 844.1 s 1.4 MB Animate baseline (30 steps, no LoRA)
fb24d48c wan22-animate-bf16 101.6 s 2.3 MB Animate + Lightx2v 4-step β€” 8.3Γ— speedup
4b4d9173 wan22-i2v-fp8 405.9 s 3.9 MB I2V cold VRAM start

All four were generated 2026-05-13 on the SceneMachine development rig.

Sister repos

πŸ€– Manifest generated by Claude Code.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support