Create and review a three-shot short drama
Connect original storyboards, short image-to-video tests, narration and editing through shot-continuity checks.
Models generate and reason. Agents use tools to work on tasks. A harness manages an agent’s environment. MCP connects tools and data. Skills are reusable task guidance.
Current task
Deliver a three-shot rough cut, shot list, three generation records, rights inventory and timecoded issue list. No missing files; inspect at least start/middle/end frames of each shot, reopen the playable export and label unresolved issues.
Prerequisites
Prepare an original three-shot story, authorized character references, ComfyUI and FFmpeg. Choose one exact official Wan2.2 template and check its model, VAE, text encoder, VRAM and license; do not mix 5B and 14B files.
Estimated time
Plan 3–5 sessions of 30–60 minutes; downloads and rendering are additional, not measured.
Validation
Original exercises based on reviewed official documentation; no model, device or regional runtime test performed.
Validation
Deliver a three-shot rough cut, shot list, three generation records, rights inventory and timecoded issue list. No missing files; inspect at least start/middle/end frames of each shot, reopen the playable export and label unresolved issues.
Limits & boundaries
Template/model mismatch, invalid frame counts, VRAM exhaustion, long-action drift, costume changes and narration desynchronization; a larger model cannot replace storyboarding and frame review.
Alternatives
comfyui
ffmpeg
whisper
Cost
Not verified
Amounts unverified. Local software can use existing hardware; model licensing, VRAM, storage, electricity and labor are separate. Cloud, APIs or subscriptions may charge.
Implementation steps
- 01For each of three shots, specify subject, action, camera, start/end state and target duration. Fix clothing and props in a character sheet; record sources and allowed uses for images, script, music and voices.
- 02Within the official template’s supported dimensions and frame counts, generate one simple action first. Save input, prompt, seed, model-file IDs and workflow. Inspect endpoints, hands/faces and background frame by frame; redesign the shot or reduce load before expanding.
- 03Generate remaining shots only after the sample passes. Rough-cut self-recorded or authorized narration with separate audio and subtitles; check character/prop/eyeline continuity and audio/subtitle synchronization. Watch the full export and retain project plus failed clips.
Limits & boundaries
- Original exercises based on reviewed official documentation; no model, device or regional runtime test performed.
- Manage script, character, image, voice and music rights separately; use fictional characters and self-recorded/preset voices without real-person permission. Disclose synthetic content without posing as a person, official source or news record.
Permissions & risks
- Manage script, character, image, voice and music rights separately; use fictional characters and self-recorded/preset voices without real-person permission. Disclose synthetic content without posing as a person, official source or news record.
Original sources
- ComfyUI native Wan2.2 video tutorial ↗
docs.comfy.org · 2026-10-16 · Official documentation
- FFmpeg official documentation ↗
ffmpeg.org · 2026-10-15 · Official documentation
- Whisper official documentation ↗
github.com · 2026-10-15 · Official documentation