Film and animation previs with ComfyUI: plan shots around documented clip limits and keep a provenance record
Teams using ComfyUI and Wan models for previsualization and animatics need shot plans that fit what the model cards state, a clear handoff format for editorial, and per-clip provenance. This guide sets out an editorial procedure from the cited official pages.
Purpose and scope
This guide helps film, TV and animation teams plan previsualization and animatic work in ComfyUI around documented model limits, editing handoff and records. It is not a tutorial and does not rate output quality.
Source-supported context
The ComfyUI Wan2.2 tutorial lists a 5B hybrid text/image-to-video model and 14B text-to-video, image-to-video and first-last-frame models; it gives no fps or clip-duration limit, though it exposes size and frame-count (length) settings and suggests a resolution around 720P for the first-last-frame example Source 1. The official Wan2.2 repository lists 480P and 720P for the A14B image-to-video model with aspect ratio following the input image, 1280x704 or 704x1280 for TI2V-5B at 24 FPS, and describes TI2V-5B as able to generate a 5-second 720P video in under 9 minutes on a single consumer-grade GPU, which is a speed claim, not a stated default clip length Source 4. The I2V-A14B card says its Mixture-of-Experts design gives reduced unrealistic camera movements, and states Apache 2.0 licensing with a usage clause that makes the user fully accountable and says use must not involve sharing content that violates laws, harms individuals or groups, spreads misinformation or targets vulnerable populations Source 2. The TI2V-5B card repeats the 720P, 24 FPS and 5-second description, adds that its 720P size is 1280704 or 7041280, and carries the same Apache 2.0 and accountability clause Source 3.
Before you start
1. A shot list with durations.
2. Source stills with known origin and model files with recorded revisions.
3. An editing destination with its expected frame rate and file type.
Steps and checks
The sequence below is an editorial procedure based on the cited context, not a diagnosis already confirmed for your environment.
Step 1. Write the shot list first: shot id, purpose, camera note, intended duration and whether the shot needs a single clip or several joined ones. Mark shots needing continuity beyond one clip.
Step 2. Fix the source frame for each shot as a saved still, and record its origin: drawn by an artist, photographed, or generated earlier, with the workflow file and the model identity. The image-to-video route uses the input image as the starting frame Source 1.
Step 3. Write down the resolution the card or README states for the model you chose next to each shot, and record the frame count (length) you set in the workflow, since the cited sources give no default clip length.
Step 4. Treat consistency as something you check, not something a model supplies. For each generated clip, compare the first frame with the source still and the last frame with the next shot's expected start; record deviations.
Step 5. Decide the handoff format per shot. Save Image writes PNG files with workflow information embedded when metadata is enabled, while Save Video writes a container file with metadata unless disabled at startup Source 6 Source 5. Ask the editor which form and frame rate the timeline expects.
Step 6. If you use VideoHelperSuite, record that it is an optional custom node and which version you installed. Its README documents a Load Image Sequence node, a frame_rate setting that changes playback duration, and a save_metadata option that includes the workflow in the output video Source 7.
Step 7. Keep a provenance sheet per delivered clip: shot id, source still, workflow file, model file and revision, licence text consulted with its date, output filename and approver.
Completion check
Every shot in the animatic has a provenance row, a stated source frame, a recorded model identity and licence date, and an agreed delivery format.
Limitations and cautions
The ComfyUI tutorial gives no duration or fps limits, so resolution and fps statements come from the model cards and repository only, and they may change with new releases. Licence text covers the model weights; it does not settle rights in your input stills, character likenesses, union or studio policies, or your jurisdiction's rules. This is not legal advice. The sources do not describe continuity across clips. VideoHelperSuite is a third-party node.
Original sources
- ComfyUI Wan2.2 tutorial — checked 2026-10-06.
- Wan2.2-I2V-A14B model card — checked 2026-10-06.
- Wan2.2-TI2V-5B model card — checked 2026-10-06.
- Wan2.2 GitHub README — checked 2026-10-06.
- ComfyUI Save Video node — checked 2026-10-06.
- ComfyUI Save Image node — checked 2026-10-06.
- ComfyUI-VideoHelperSuite README — checked 2026-10-06.
Prepared 2026-10-06 from the sources above. No GPU run, quality measurement or runtime guarantee is supplied.
Related troubleshooting and guides
- Wan 2.2 5B versus 14B: match the task, model stages and VAE
- First Wan 2.2 5B image-to-video task: verify the input through to the saved clip
- VideoHelperSuite reference: frame loading, audio, composition and saved output
- Record model downloads by repository, revision, path and integrity evidence
- Retrieve ComfyUI API outputs from task history and /view
Was this page helpful?
Anonymous. Only a yes/no counter is stored; no account, IP address or device details.
Sources & references
Editorial research from the cited public sources, dated 2026-10-06. This site did not run or test the workflow described.
01ComfyUI Wan2.2 tutorialSource checked: 2026-10-0602Wan2.2-I2V-A14B model cardSource checked: 2026-10-0603Wan2.2-TI2V-5B model cardSource checked: 2026-10-0604Wan2.2 GitHub READMESource checked: 2026-10-0605ComfyUI Save Video nodeSource checked: 2026-10-0606ComfyUI Save Image nodeSource checked: 2026-10-0607ComfyUI-VideoHelperSuite READMESource checked: 2026-10-06Report an issue · 182dd84d-2bc1-5845-a77f-80f23a511a68
Related reading
Pages linked to this one by the editors.
Distinguish 5B TI2V from the 14B text/image-to-video routes. A shared Wan 2.2 label does not make weights, VAE, encoder and latents freely interchangeable.
- Wan 2.2
This guide follows the 5B TI2V image-input path specifically. The model-family reference explains alternatives; this page handles input, execution stages and output verification.
- Wan 2.2
VHS_VideoCombine illustrates the media layer around generated content. Reading, sampling frames, combining and saving should not be confused with the generation model itself.
A model name can cover bundled checkpoints, split weights, quantized variants, text encoders and VAEs. Preserve enough identity information to know which artifact a workflow actually used.