The axcel-film-studio skill: the step-by-step way to make a multi-scene AI film or story video with Axcel (owner decisions first, including landscape or portrait; story; reference images; scenes; download and QA; HyperFrames edit with HeyGen or ElevenLabs sound; mastering; subtitles). Call it before starting a film when your client has no skills support. Omit section for the main guide.

What this Axcel MCP can do, the recommended workflow, and example requests.

Whether you are signed in, the account plan, credits and limits, plus local settings. Call this first.

Open a browser window on axcel.cloud so the user can sign in with their own Google account. Returns immediately. Then tell the user to sign in there, wait for their reply, and call axcel_login_complete.

Confirm the sign-in started with axcel_login_begin and save the session (waits briefly while the window is open).

Screenshot and text of every page of the open sign-in window (including the Google popup), to see what the user sees when sign-in is stuck.

Fallback sign-in when the browser window cannot be used. The user saves their axcel.cloud cookies to a local file (DevTools > Network > any /api request > Request Headers > cookie, or a cookie-extension JSON export, or cookies.txt). Never ask for cookie values in chat; only for the file path.

Forget the saved Axcel session on this computer (and by default the sign-in browser profile).

Live model catalog with the options each model accepts: inputs (text/start/end/refimg/...), resolutions, durations, ratios and audio.

Read-only list of Axcel plans and prices. This MCP never buys anything.

Generate a video on Axcel (text-to-video or image-to-video), wait for it and download it. Returns a jobId; if it is still processing, poll axcel_job_status with that jobId.

Generate an image on Axcel (e.g. Nano Banana Pro, Seedream 5.0 Pro), optionally guided by up to 4 reference images, wait for it and download it.

Status of an Axcel job by jobId. Waits up to waitSeconds for it to finish and downloads the result when done. Use the SAME jobId; never resubmit a slow job.

Jobs currently running on the Axcel account.

Download a finished job (video or image) with a .axcel.json manifest (SHA-256, media info).

Return a small preview image of a local image or video (one frame) or of a finished job, so you can check the result visually. Needs ffmpeg for videos.

Upload a local image (to the Library) or a reference video/audio, and get its assetId for reuse in later generations (kept 14 days).

Generation history on Axcel (kept 7 days), newest first. Use the returned cursor for the next page.

One history item, including its result URL.

Delete a history item on Axcel (cannot be undone).

Library items (uploaded images and saved results; kept 14 days). Use the returned "berikut" value as before for the next page.

Save a finished job result into the Library, optionally into a folder.

Delete a Library item (cannot be undone).

List Library folders.

Create a Library folder.

Rename a Library folder.

Delete a Library folder (cannot be undone).

Move Library assets into a folder (folderId null = out of any folder).

Download a whole Library folder as one ZIP file.

Save one frame (last, first or at N seconds) of a local video or a finished job as PNG, e.g. to start the next scene from it. Needs ffmpeg.

Join clips into one MP4 (normalised size/fps/audio; optional trim of the repeated first frame and crossfades). Needs ffmpeg.

Generate scenes one after another in the background. With chain=true each scene starts from the previous scene's last frame; at the end the clips are joined into one MP4. Returns a pipelineId; poll axcel_sequence_status. Needs ffmpeg for chaining/joining.

Progress of a multi-scene film. resume=true continues a paused or stopped film where it left off (finished scenes are never resubmitted; re-running a FAILED scene needs confirm=true).

Multi-scene films started from this computer, newest first.