qwen-mm-plugins-omni-chatcut-music-to-mv

Provide the source audio and a project directory; the Skill runs music captioning to extract structure and lyrics, authors a validated timed storyboard, generates one video segment per editorial shot via Qwen Image/Wan or Seedream/Seedance, then assembles the final MV with technical QC.

Run the state inspection script on the project directory. If resumable execution state or an incomplete caption run with readable source audio exists, the Skill continues from the last durable phase instead of restarting.

It defaults to Qwen Image 3.0 and Wan 3.0 on DashScope using a Beijing-region DASHSCOPE_API_KEY. Seedream and Seedance connect directly to Volcengine Ark, with Seedance using existing Ark image asset IDs for fixed characters.

No. Omni semantic shot QC is off unless the user explicitly requests it or sets quality_control.enabled=true in the run config. When skipped, the Skill reports it as skipped without treating it as a required stage.

Execution requires a validator-approved storyboard with creative direction, one provider request per editorial shot, and a passing continuity report. Failed validation returns work to storyboard authoring because structural problems cannot be fixed during generation.

Music analysis, validation, and planning are free, but an explicit request to create the final MV or resume generation authorizes provider calls. If intent is ambiguous, the Skill stops before the first paid call and asks once.