PixVerse CLI
Generate media using the user's PixVerse account and credits. CLI access requires a subscription. Use --json for machine-readable results.
Read only what the task needs
This is the entrypoint; the linked Markdown files are supporting instructions, not separately installed commands. Resolve links relative to this file, regardless of the working directory. Select one capability or workflow below; follow additional links only for a needed operation. Do not load the entire catalog, every related file, or all design references.
For a single operation, read its capability. For a multi-step deliverable, start with the workflow. Before executing a creation chain, read the execution contract once for output shapes, waiting, and recovery. For prompt advice or rewriting, select one strategy using the prompt contract.
Setup and parameter discovery
- Check
pixverse --version. If missing, install withnpm install -g pixverse(Node.js >= 22.12); an existing older installation need not be changed for unrelated tasks. - On CLI 1.4.0+, query only the required mode/model:
pixverse capabilities create video --model v6 --json. Usepixverse capabilities create --jsonto discover mode/model IDs if unknown. These queries are offline and need no login. Avoid the fullpixverse capabilities --jsonbundle for a narrow task. - On older versions, use
pixverse create <mode> --helpand the relevant capability's static reference. This skill documents CLI 1.4.15; a static table does not prove an older installation supports a model or flag. If support cannot be established, report the required upgrade instead of submitting an invented combination. - Before account operations or generation, check
pixverse auth status --jsonif login state is unknown. ForPIXVERSE_AUTH_PROFILE=chatgpt, use MCP pairing instead of ordinary login; retain this profile throughout the workflow. For ordinary login, runpixverse auth login --jsonand show the returned authorization URL; JSON mode does not open the browser automatically. See auth and account for access keys, configuration, and recovery.
Canvas capabilities and MiniApp schemas are live: query the relevant node/app before constructing a request. See capability discovery. Do not infer Canvas node mappings from Create model names.
Select an operation
| Task | Read |
|---|---|
| Text/image to video; reference generation or source-video editing | create-video |
| Create or edit an image | create-and-edit-image |
| Modify a video at a keyframe | modify-video |
| Transfer motion from a video to a character | motion-control |
| Extend or upscale video | post-process-video |
| Animate between keyframes | transition |
| Speech / voiceover | create-voice |
| Music / soundtrack | create-music |
| Browse and run effects | template |
| Discover and run preset apps | miniapps |
| Build, operate, or automatically arrange a Canvas graph | canvas |
| Query/wait for an existing task | task-management |
| Upload, inspect, download, or delete assets | asset-management |
| Organize saved folders | saved-folders |
| Login, credits, usage, or creation defaults | auth-and-account |
| List or select workspaces | workspace |
Select a creative strategy
Explicit user constraints take priority over default creative advice. An instruction to optimize and generate authorizes that rewrite; an ordinary generation request does not authorize silently changing a supplied prompt. Do not run several optimizers sequentially.
| Task | Read |
|---|---|
| General prompt advice or models without a dedicated optimizer | prompting-guide |
| V6 prompt rewriting | prompt-enhance |
| Seedance prompt structure, references, editing, or shot control | seedance-prompt-optimize |
| Seedance emotional / atmospheric concept development | seedance-vibe-creating |
| Persistent characters | character-design |
| Persistent props / items | item-design |
| Mondo-style poster, cover, or album artwork | mondo-poster-design |
Multi-step deliverables
| Task | Read |
|---|---|
| Text to finished video | text-to-video-pipeline |
| Animate an existing image | image-to-video-pipeline |
| Generate an image, then animate it | text-to-image-to-video |
| Iterative image edits | image-editing-pipeline |
| Modify, then enhance video | modify-video-pipeline |
| Motion transfer with post-processing | motion-control-pipeline |
| Video with optional extension, upscale, and soundtrack | video-production |
| Parallel or multi-output generation | batch-creation |
| Mondo poster production | mondo-poster-pipeline |
| Animate a Mondo poster | mondo-poster-to-video-pipeline |
| Four-shot storyboard and final edit | storyboard-to-video |
Execution essentials
- Create normally waits for completion. Save its result; do not create again to extract an ID or wait again for a completed result. For background work use
--no-wait, retain all IDs, then wait/query existing tasks. - Preserve the requested model, region, workspace, and creative constraints. Query supported parameters before selecting replacements. Text inputs (
--prompt,--text,--lyrics) accept literal text, a file path, or-for stdin. PIXVERSE_REGIONoverrides--region; default isglobal. Login, active workspace, and creation defaults are isolated by region. CN does not support standalone voice/music or audio asset/task operations; model availability also differs. See the selected capability before submission.--workspace-id <id>overrides workspace for one invocation;0is personal. If the CLI resets an inaccessible active workspace to personal, re-establish the intended destination before retrying team work.- Preserve successful outputs on partial failure. A nonzero exit can accompany usable batch results. A timeout is not proof of a failed submission. Follow the execution contract, not a blind create retry.
For Windows shell syntax, see the PowerShell pipeline. For updating this skill checkout, run update.sh only from a clean main checkout; it supports fast-forward updates and does not stash or switch branches.
