Wan2GP Operator

Codex-native command center for Wan2GP.

Stop babysitting sliders. Stop guessing memory limits. Stop burning runs on broken settings. wan2gp-operator turns Wan2GP into a disciplined, terminal-first production workflow.

Why This Exists

Wan2GP is powerful. But raw power without guardrails is chaos:

• fragile prompts
• wrong runtime flags
• bad checkpoint states
• wasted generations
• no consistent troubleshooting loop

Wan2GP Operator adds the missing layer: operational intelligence.

Why This Beats Using Wan2GP Directly

Using Wan2GP directly:

• manual setup drift
• easy to misconfigure runs
• no standard dry-run workflow
• no persistent learning from failures
• weak postmortem trail

Using Wan2GP Operator:

• guided install + readiness checks
• VRAM-aware compose flow
• deterministic run --dry-run -> run pipeline
• structured logs + diagnostics
• evolving compatibility/quality state over time

This is not "more UI." It is better control.

Core Capabilities

• Bootstrap install and launch
• Hardware/readiness assessment
• Prompt-to-settings composer
• Headless batch execution
• Auto-retry for known CLI incompatibilities
• Failure diagnosis with next-step commands
• Upstream release tracking
• Recursive evolution state (.wan2gp_operator_state.json)
• Music-video pipeline (audio analyze -> beat plan -> clip generation -> ffmpeg assembly)

Workflow

python scripts/wan2gp_operator.py bootstrap
python scripts/wan2gp_operator.py compose --prompt "cinematic street shot" --quality quality --duration-seconds 4
python scripts/wan2gp_operator.py run --wan-root <WAN2GP_ROOT> --process <settings.json> --dry-run
python scripts/wan2gp_operator.py run --wan-root <WAN2GP_ROOT> --process <settings.json> --log-file logs/run.log
python scripts/wan2gp_operator.py evolve --wan-root <WAN2GP_ROOT> --log-file logs/run.log

Music Video Pipeline (Phase 1)

Turn one track into a structured multi-shot video using Wan2GP as the generation engine.

End-to-end

python scripts/wan2gp_operator.py music-video \
  --audio "<SONG>.mp3" \
  --theme "neon summer city, confident performance energy" \
  --wan-root <WAN2GP_ROOT> \
  --execute-generation \
  --evolve-on-failure

Stage-by-stage

python scripts/wan2gp_operator.py music-analyze --audio "<SONG>.mp3"
python scripts/wan2gp_operator.py music-plan --analysis <audio_analysis.json> --theme "<THEME>"
python scripts/wan2gp_operator.py music-generate --plan <music_plan.json> --wan-root <WAN2GP_ROOT> --execute-generation
python scripts/wan2gp_operator.py music-assemble --audio "<SONG>.mp3" --manifest <generation_manifest.json>

Generated artifacts

• audio_analysis.json: duration, BPM, beat grid, sections
• music_video_plan.json: beat-aligned shot timeline and prompts
• generation_manifest.json: take-by-take compose/run/evolve results
• music_video_master.mp4: assembled output with track muxed in

Dependencies

• Required: ffmpeg, ffprobe
• Optional but recommended: librosa (better beat detection)

Skill Layout

• SKILL.md: skill contract and routing
• scripts/: operational CLI tools
• references/: practical runbooks and tuning notes
• agents/: agent-facing metadata

Install As A Codex Skill

Copy this folder into your Codex skills directory as wan2gp-operator.

Windows example:

Copy-Item -Path ".\\wan2gp-operator-skill" -Destination "$env:USERPROFILE\\.codex\\skills\\wan2gp-operator" -Recurse -Force

Then restart Codex and invoke the skill by name.

Positioning

Wan2GP Operator is for builders who care about outcomes, not ritual. You can keep clicking around and hoping for magic. Or you can run a repeatable video ops pipeline that gets better every time.