Skip to content
SupportOpen app

Workflows and music videos

Use these tools to run workflows and to cut a video to a music track.

Not every account has these tools. You build your own workflows in the Screel editor, and Screel ships built-in templates. These tools find both, run them, follow the run and answer the steps that pause for a person. They do not change a workflow’s steps.

The workflows of a project, or of your personal library when no project is named. Each one comes with its Run form: the param steps with their name, kind, default and options. Also lists the built-in templates under templates, each with its key, title, description and Run form. It spends no credits.

  • Surfaces: Generic MCP, Manan
  • Parameters: project_id

Start a run with the Run form values and a credit budget; without a budget the run gets the default one. Name one of your workflows with document_id, or a built-in template with template and a project. The first run puts the template into the project as a workflow, and later runs reuse that copy, edits included. A library workflow run with a project runs as a copy in that project. A picture, video or audio input takes a public https link to the file, or the id of a file Screel already holds. Screel stores a linked file in your project first. A budget below the run’s estimate starts nothing, and a run that reaches its budget pauses instead of going over. It returns the run’s job_id at once, and the run continues in the background. It spends credits up to the budget.

  • Surfaces: Generic MCP, Manan
  • Parameters: budget_credits, document_id, params, project_id, template

One run’s status, outcome, credits and budget, and every step’s status, error and a short summary of its output. A paused step also lists its question, its options and its candidates, each with the index a choice takes. An agent step’s question can also list suggestions, short answers the agent proposes, and question_read, one line on what it concluded; any other answer stays valid. It spends no credits.

  • Surfaces: Generic MCP, Manan
  • Parameters: job_id (required)

Answer a paused step: pick one of its options or one candidate, pass every candidate on, or answer its question. A run paused at its budget continues only in the Screel editor. The tool spends no credits, and the steps after it spend within the run’s budget.

  • Surfaces: Generic MCP, Manan
  • Parameters: job_id (required), node_id (required), answer, choice

These tools cut a video to the project’s music track. The scenes change on the beats, and the track plays as the soundtrack. On ChatGPT app v2, the song comes from generate_music. You can also add your own track as the project’s music in the Screel web app, because no tool there takes an audio file. In Manan, attach the track to your message and ask for a music video.

Read a music track for a video cut to the music: its tempo, beats, sections and timed lyric lines. Each section reports its energy and whether a voice sings. Every time is in seconds of track time. With no track named, the tool reads the project’s music track. It also takes the id of an audio file Screel holds. On Generic MCP and in Manan, it takes a public https link (audio_url), and Screel stores the file in your project first.

Pass the lyrics when you know them, and Screel times the lines instead of writing them down from the vocals. Screel keeps the analysis with the track, so a second read answers at once and spends nothing. plan_music_cuts and a workflow’s Music cuts step read the same analysis. A new analysis returns a job_id, and get_job_status carries the analysis when the job finishes.

A new analysis spends credits, and a repeat is free.

  • Surfaces: ChatGPT app v2, Generic MCP, Manan
  • Parameters: project_id (required), audio_url, lyrics, request_id, resource_id

Plan a music video’s scenes from the analysis of the project’s music track. Each slot is one scene with its start and length on the beats. A slot also names its role: a still, a clip, or a clip of the singer. It names its song section, the section’s energy and the lyric heard in it. mode picks story, performance or lyric. Story holds stills on the beats and makes clips in the chorus and the drop. Performance makes clips wherever a voice sings. Lyric cuts and captions at every lyric line.

start and end pick an excerpt, and max_scenes caps the scenes (48 by default, 90 at most). Each clip slot fits the lengths a clip can have, so animating it keeps the cut. A part too short or too long for a clip stays a still, so a low max_scenes can leave the plan with no clips. A plan with a scene longer than 30 seconds fails: raise max_scenes or pick a shorter excerpt.

Without apply, the tool changes nothing, and its result says how to retime the scenes a project already has. With apply, the tool writes the scenes into a project that has none yet. It also makes the track the music master: the video plays the song once at full volume and mutes the clips’ own sound. The result then gives the credits that animating the clip scenes takes, for each scene and in total.

In Manan and on ChatGPT app v2, it also gives the credits of the same scenes as lip-sync clips. The tool itself spends no credits.

  • Surfaces: ChatGPT app v2, Generic MCP, Manan
  • Parameters: project_id (required), apply, end, max_scenes, mode, start