agentleFS
vllm-projectvllm-project/vllm-omni/.agents/skills/vllm-omni-h3-character-replacement/SKILL.md

vllm-omni-h3-character-replacement

Skill7.8k starsChanged 3 days agoApache-2.0View on GitHub

Plan MiniMax-H3 character replacement, write per-segment Ref2VA prompts and configure reusable H3-04 ComfyUI workflows for generation and review.

Adopted by

Steady: 0 new projects took it in the last 4 weeks, none the 4 weeks before.thin data

Week of Sep 20: 0 new projectsWeek of Sep 27: 0 new projectsWeek of Oct 4: 0 new projectsWeek of Oct 11: 0 new projectsWeek of Oct 18: 0 new projectsWeek of Oct 25: 0 new projectsWeek of Nov 1: 0 new projectsWeek of Nov 8: 0 new projects
Sep 2026Oct 2026Nov 2026
New projects taking a version of this file each week, over the 8 weeks our latest run read: only 0 in all, so this is a hint, not a trend.

Who's talking about vllm-project/vllm-omni

The biggest 5 of 51 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 51

Talked about on GitHub and Hacker News (51 times).

On GitHub · 50

45 more on GitHub

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read prompt-writing.md, then write each segment's prompt from its inspected action and explicit source-person-to-picture mappings.

    line 13
  2. Distinctive:

    When generation is requested, submit the executable graph through ComfyUI, monitor results, inspect segment previews and deliver the …

    line 15

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

vLLM-Omni H3 Character Replacement

Make character replacement easy to use: the user supplies source media and a replacement request; the agent derives the segment plan, prompts and workflow parameters. Keep the input boundary as original images, source video and explicit prompts. Work only on the requested deliverable; planning or prompt edits do not start generation, restart services or introduce preprocessing.

For a simple one-person edit without segmented continuation, reuse the H3-04 Character Replacement workflow. For planned multi-segment work, use the Agentic Continuation template linked below; its example-specific settings must be recalculated for new footage.

  1. Create a 24-FPS constant-frame-rate working copy of the source video before visual inspection, segmentation and frame-index calculations, unless the user explicitly requests another FPS. Preserve the original file and playback timing. Read video-segment.md first. Use a vision-capable model to inspect the source video, establish shot boundaries and track when each selected person appears, interacts, becomes occluded and leaves. Produce the segment plan before writing new segment prompts.
  2. Read prompt-writing.md, then write each segment's prompt from its inspected action and explicit source-person-to-picture mappings.
  3. For a single-person, single-segment edit, configure the simple workflow linked above. For segmented or multi-person work, configure a task-specific copy of the H3-04 Agentic Continuation workflow, reusing Global Control, existing nodes and one shared segment-subgraph definition. Follow workflow.md for configuration, execution and review.
  4. When generation is requested, submit the executable graph through ComfyUI, monitor results, inspect segment previews and deliver the assembled video with remaining quality defects.

Prompt-format dependency: read the official MiniMax H3 prompt-writing skill. It can be read directly from the repository; local installation is optional. Before execution, establish the ComfyUI URL, installed node availability, effective Omni deployment settings and task output directory as described in workflow.md.

Resolve repository-relative references from this skill directory after resolving any global skill symlink, not from the agent's current working directory.

Execution boundary: agent → ComfyUI execution API → existing vLLM-Omni nodes → vLLM-Omni video API. The nodes own model requests; the agent does not bypass ComfyUI with a separate generation client.

For a prompt-only correction to an existing segment, inspect its relevant footage and existing plan; revise the plan only if the observed defect exposes an incorrect boundary or mapping. Do not invent a new segmentation for every wording edit.

Preserve the user's model, resolution, FPS and fixed seed unless asked to change them. Report written plans and prompts separately from generation and visual acceptance. Keep Markdown paragraphs on single source lines and usage notes brief.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.