Skills

Skill 25 of 34

Generate Video

Create videos with RunComfy AI video models including Seedance, Wan, LTX and Kling.

2 minutes · 423 words · 4 sections

Install

$ npx skills add runcomfy-com/skills --skill generate-video

$ npx skills add runcomfy-com/skills

» /plugin marketplace add runcomfy-com/skills

The first command installs just this skill, by the name in its SKILL.md; the second installs the whole repository.

Generate videos with RunComfy

Use the connected RunComfy MCP tools; tool names may have a client-specific prefix. If authentication is needed, use the client’s RunComfy sign-in flow. Keep credentials out of chat and files.

Plan the shot

Reuse the user’s subject, action, camera motion, duration, aspect ratio, resolution, audio preference and budget. Clarify only the details needed to choose a valid request.
Search list_models by the requested family and capability. RunComfy includes Seedance 2.5, Wan 3.0 and LTX 2.5, as well as Kling models; current availability comes from the returned model list. Use list_model_categories when needed.
Call get_model for the exact returned model_id. Distinguish text-to-video, image-to-video, reference-to-video, first/last-frame and editing endpoints. Do not infer supported inputs from a family name.
Read the exact required input names and limits. Match duration, resolution, aspect ratio, audio and seed to the schema. A pricing tier label may differ from the encoded output dimensions.
Check get_balance. Show the complete request and available price estimate, including per-second pricing when applicable. Use existing user authorization and spending limits; obtain missing authorization before paid submission or publication of private input assets.

Submit once and monitor

Use publicly accessible HTTPS URLs for model file inputs. Do not send local paths or private browser links. If the user provides a local file, use an available authorized upload route; otherwise explain the requirement.
Call run_model once with wait_for_completion: false. Save its request ID and use get_model_request_status to monitor that same job.
Poll with reasonable spacing. Do not resubmit just because generation is slow, the client times out, or the response is interrupted. Reconcile the existing request first.
Fetch get_model_request_result after completion. Use cancel_model_request if the user cancels or an agreed stopping condition is reached.
Inspect the cancellation outcome: an in-progress or finished request may be not_cancellable. Report that result and continue reconciling the job. A monitoring deadline does not guarantee cancellation or a hard spending cap, and already-incurred charges remain possible.
Return the actual video URL and request ID. Open or embed it if the client supports video. When inline playback fails, provide the exact direct output link.
Distinguish a completed job, successful file retrieval and actual playback verification. Report observed duration, resolution and audio only if inspected; avoid claiming visual quality from status alone.

Useful requests

“Compare Seedance, Wan and LTX for a five-second product clip. Show current prices and supported inputs; do not run yet.”
“Animate this mug image with a slow camera orbit. Review the exact model, duration and cost first.”

AI video models (opens in a new tab) · Model API documentation (opens in a new tab)

About this skill

Trigger

Create videos with RunComfy AI video models including Seedance, Wan, LTX and Kling. Use for text-to-video, image-to-video, reference-to-video and video edits. Check model-specific duration, aspect ratio, audio, inputs and price before submitting one job.

The verbatim description from this skill’s front matter — the string an agent matches on to decide whether to load it.

Colophon for this issue

Colophon

How this issue was made

Source

github.com/runcomfy-com/skills, branch main, last pushed 15 September 2026.

Licence

MIT — the text of every skill is reproduced unmodified, frontmatter included, under the upstream licence.

Discovery

34 skills found by walking the repository tree for SKILL.md, not by matching a directory convention. 2 distinct layouts observed: */SKILL.md, plugins/runcomfy/skills/*/SKILL.md.

Issue colours

Resolved from a deterministic hash of the owner name. Two accent tones are generated per issue and each is proven against its own ground before it ships: a single accent that passes AA on both light and dark paper is arithmetically impossible.

Heading repairs

1 repair applied to this skill so the document has one h1 and no skipped levels:

Shifted “4 headings” from h1 to h2 so the skill title is the only h1.

Spec compliance

2 editorial notes across 2 of 34 skills. They are printed in the margin of each skill rather than as errors here.

AI Music — 1 note
Runcomfy CLI — 1 note

Images

Images inside a skill come from the upstream repository. Where the author gave no alternative text we mark the image decorative rather than inventing a description — a plausible caption we made up is worse than none for the reader who depends on it.

Marketplace

A plugin manifest is published at .claude-plugin/marketplace.json by RunComfy, declaring 1 plugin. It is read for editorial metadata only — never as the skill index, which is always the repository tree.

Signal

This repository is unranked on skills.sh, which says nothing about it: the leaderboard covers a fraction of published skills.

Agent surfaces

The whole issue is available as one markdown document at /runcomfy-com/skills.md, and each skill at its own .md URL.

Publication

Set by Skills Docs from the source repository. Body text is Literata at the reader’s chosen size and measure; code is Geist Mono. Nothing on this page was written by us except this paragraph.

Generate videos with RunComfy

Use the connected RunComfy MCP tools; tool names may have a client-specific prefix. If authentication is needed, use the client’s RunComfy sign-in flow. Keep credentials out of chat and files.

Reuse the user’s subject, action, camera motion, duration, aspect ratio, resolution, audio preference and budget. Clarify only the details needed to choose a valid request.

Search list_models by the requested family and capability. RunComfy includes Seedance 2.5, Wan 3.0 and LTX 2.5, as well as Kling models; current availability comes from the returned model list. Use list_model_categories when needed.

Call get_model for the exact returned model_id. Distinguish text-to-video, image-to-video, reference-to-video, first/last-frame and editing endpoints. Do not infer supported inputs from a family name.

Read the exact required input names and limits. Match duration, resolution, aspect ratio, audio and seed to the schema. A pricing tier label may differ from the encoded output dimensions.

Check get_balance. Show the complete request and available price estimate, including per-second pricing when applicable. Use existing user authorization and spending limits; obtain missing authorization before paid submission or publication of private input assets.

Use publicly accessible HTTPS URLs for model file inputs. Do not send local paths or private browser links. If the user provides a local file, use an available authorized upload route; otherwise explain the requirement.

Call run_model once with wait_for_completion: false. Save its request ID and use get_model_request_status to monitor that same job.

Poll with reasonable spacing. Do not resubmit just because generation is slow, the client times out, or the response is interrupted. Reconcile the existing request first.

Fetch get_model_request_result after completion. Use cancel_model_request if the user cancels or an agreed stopping condition is reached.

Inspect the cancellation outcome: an in-progress or finished request may be not_cancellable. Report that result and continue reconciling the job. A monitoring deadline does not guarantee cancellation or a hard spending cap, and already-incurred charges remain possible.

Return the actual video URL and request ID. Open or embed it if the client supports video. When inline playback fails, provide the exact direct output link.

Distinguish a completed job, successful file retrieval and actual playback verification. Report observed duration, resolution and audio only if inspected; avoid claiming visual quality from status alone.

“Compare Seedance, Wan and LTX for a five-second product clip. Show current prices and supported inputs; do not run yet.”

“Animate this mug image with a slow camera orbit. Review the exact model, duration and cost first.”