Skip to content
Back to skills

Latent Sync

ASecurity

"Route LatentSync video generation, raw-video preprocessing,

  • 247 stars
  • 0 votes
  • 0 copies
  • 3 views
  • Added September 8, 2026
toolspythonapibackend

Works with

  • cli
  • api

Security analysis

A100/100

Pro scans all 20 files and shows the line behind each finding

Scanned September 8, 2026

npx -y skills add VectorSpaceLab/AREX-Skill --skill latent-sync --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Latent Sync?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Latent Sync
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vectorspacelab-latent-sync/badge)](https://www.skillsdirectory.com/skills/vectorspacelab-latent-sync)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: latent-sync
description: "Route LatentSync video generation, raw-video preprocessing,
  training, and evaluation workflows to the right sub-skill."
disable-model-invocation: true
metadata:
  disco-role: operating
license: Apache 2.0
---

# LatentSync

Use this skill when the request is about the LatentSync repository: talking-head video synthesis, raw `.mp4` preprocessing, U-Net or SyncNet training, or SyncNet/FVD evaluation.

## Quick start

1. Run `python scripts/check_env.py --check-imports --check-cuda --check-ffmpeg` before a long workflow.
2. Add `--check-scenedetect` when preprocessing raw clips.
3. Add `--check-assets` when you want the bundled demo inputs or Gradio smoke path.
4. Read the narrow sub-skill before choosing configs or launch commands.

## Route by task

- Generate one video, a small batch, or launch the local UI: [`sub-skills/inference/SKILL.md`](sub-skills/inference/SKILL.md).
- Convert raw videos into aligned, AV-synced, quality-filtered clips: [`sub-skills/data-preparation/SKILL.md`](sub-skills/data-preparation/SKILL.md).
- Choose U-Net or SyncNet configs, build fileslists, or render training launches: [`sub-skills/training/SKILL.md`](sub-skills/training/SKILL.md).
- Score generated clips, validate SyncNet checkpoints, or compare real and fake sets: [`sub-skills/evaluation/SKILL.md`](sub-skills/evaluation/SKILL.md).

## Shared references and helpers

- [`references/configuration.md`](references/configuration.md) — repo-wide config, checkpoint, and resolution map.
- [`references/api-reference.md`](references/api-reference.md) — verified shared runtime call surfaces.
- [`references/troubleshooting.md`](references/troubleshooting.md) — cross-cutting install, import, backend, path, and checkpoint failures.
- [`references/repo-provenance.md`](references/repo-provenance.md) — source snapshot and refresh baseline.
- [`references/repo-routing-metadata.json`](references/repo-routing-metadata.json) — structured router metadata for managed discovery.
- [`scripts/check_env.py`](scripts/check_env.py) — safe repo-root, import, CUDA, ffmpeg, scenedetect, and asset checker.

## Working rules

- Always pass `--repo-root` to bundled helpers when the current directory is not the runtime tree.
- Keep temp directories isolated; the pipeline and metrics delete and recreate scratch paths.
- Treat CUDA as required for the core inference, preprocessing, and training workflows; CPU-only checks are useful only for limited import or FVD-style smoke paths.
- Do not route a training, preprocessing, or metric issue through inference just because the same checkpoints or demo assets appear in the stack.
- Use `references/repo-provenance.md` when deciding whether this skill still matches the current checkout or needs refresh.

Files in this skill

  • SKILL.md2.7 KB
  • references/api-reference.md4.5 KB
  • references/configuration.md2.7 KB
  • references/repo-provenance.md2.1 KB
  • references/repo-routing-metadata.json509 B
  • references/troubleshooting.md7.7 KB
  • scripts/check_env.py12.4 KB
  • sub-skills/data-preparation/SKILL.md3.3 KB
  • sub-skills/data-preparation/references/data-formats.md7.7 KB
  • sub-skills/data-preparation/references/troubleshooting.md8.9 KB
  • sub-skills/data-preparation/references/workflows.md10 KB
  • sub-skills/data-preparation/scripts/check_data_prep_inputs.py17.8 KB
  • sub-skills/data-preparation/scripts/run_data_pipeline.py22.1 KB
  • sub-skills/evaluation/SKILL.md3.6 KB
  • sub-skills/evaluation/references/metrics.md5.7 KB
  • sub-skills/evaluation/references/troubleshooting.md6.7 KB
  • sub-skills/evaluation/references/workflows.md7.9 KB
  • sub-skills/evaluation/scripts/plot_syncnet_curves.py4.1 KB
  • sub-skills/evaluation/scripts/run_evaluation.py21.1 KB
  • sub-skills/inference/SKILL.md3 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.