Install
openclaw skills install @fuyunzhishang/pixmind-video-to-promptExtract a storyboard table with local video frames
openclaw skills install @fuyunzhishang/pixmind-video-to-promptTurn one video into a portable structured storyboard. The bundled script submits and polls Pixmind, parses the returned scene nodes, extracts reference frames locally, and emits a generic table presentation that compatible clients can render without Pixmind-specific code.
This API costs 100 credits per started minute, rounded up. It accepts one video per request and rejects videos longer than 10 minutes.
Before submitting:
ceil(seconds / 60) * 100 credits.Use one source only. Default to zh-cn and 100 scenes unless the user supplied other values.
The launchers use an existing Node.js installation when available. Otherwise they install a
portable, SHA-256-verified Node.js runtime in the Pixmind cache. Pixmind R2 is the primary
source and the matching nodejs.org release is the fallback; neither path modifies the system
PATH. On Windows use run.ps1; on macOS use run.sh.
Remote video on Windows:
powershell -NoProfile -ExecutionPolicy Bypass -File "{baseDir}/scripts/run.ps1" --url "https://example.com/video.mp4" --language zh-cn --max-scenes 100 --yes --poll
Local video on Windows:
powershell -NoProfile -ExecutionPolicy Bypass -File "{baseDir}/scripts/run.ps1" --file "C:\path\to\video.mp4" --language zh-cn --max-scenes 100 --yes --poll
Remote video on macOS:
sh "{baseDir}/scripts/run.sh" --url "https://example.com/video.mp4" --language zh-cn --max-scenes 100 --yes --poll
Local video on macOS:
sh "{baseDir}/scripts/run.sh" --file "/path/to/video.mp4" --language zh-cn --max-scenes 100 --yes --poll
Resume an existing task without resubmitting or charging again. Use the launcher for the current platform:
sh "{baseDir}/scripts/run.sh" --task-id 12345 --file "/path/to/original.mp4" --poll
Keep the original source when resuming if screenshots are required. Supplying --task-id never submits another paid task; --file or --url is used only for local frame extraction.
The script takes the midpoint of every returned splits[index], extracts a 9:16 centered JPEG, and pairs it with data[index].
--ffmpeg, then PIXMIND_FFMPEG_PATH, then system ffmpeg.result.json under ~/Downloads/pixmind-video-to-prompt/{taskId}/ unless --output is supplied.storyboard, local screenshots, attachments, the original raw response, and a generic presentation object.presentation when the host supports structured results. It contains the SaaS-equivalent sections and seven-column table without requiring client-specific Pixmind logic.attachments and render the scenes in a Markdown table with: scene number, reference image, duration, shot type, script, scene description, and video prompt.ready, success, and completed as successful terminal states.failed, error, and cancelled as unsuccessful terminal states and report the API message.PIXMIND_API_KEY; never ask the user to paste the key into chat.