Puffin Ship Docs

Starter Templates

The Gallery tab in Build a Template opens with four starter templates. Each one is a complete, working pipeline: every step already has its inputs, settings and output level filled in, so you only have to supply your footage and any service key the template needs. Pick one, and it is copied into your workspace as a normal template you can edit.

Template What it does You supply Service key
Assemble Clips Joins several clips end to end A folder of clips None
Script, Record, Caption Writes a script, records you reading it, captions the take Nothing OPENAI_API_KEY
Trim Dead Space Cuts the silences out of a long recording One video OPENAI_API_KEY
Auto-Caption a Clip Puts per-word captions on a clip One video OPENAI_API_KEY

Service keys are set once per workspace under Workspace Settings, Variables & Keys. See Getting Started for where to find them.

Assemble Clips

The simplest one, and the best first run: no AI, no service key, no waiting.

Upload the clips you want joined into a folder named clips. They are placed end to end in filename order, so name them clip-01.mp4, clip-02.mp4, clip-03.mp4 and they play in that order. Run it and you get one continuous edit you can preview and export.

Clips keep their own format, so the result looks best when they share a resolution and frame rate. If you are mixing footage from different cameras, expect the preview to change size between clips.

To change the order, either rename the files or replace the clips input with an explicit list of file names in the order you want.

Script, Record, Caption

The full loop, with nothing to upload.

  1. Script. Claude writes a short script. The step ships with a placeholder brief; replace it with a few lines about your own product. The run pauses at a review gate so you approve the wording before you say it out loud.
  2. Record. The run pauses again and opens the recording panel, using the approved script as a teleprompter. See Recording in the browser.
  3. Transcript. Your take is transcribed, so the captions match what you actually said rather than what the script planned.
  4. Captions. Per-word captions are burned on, sized for a vertical frame.

It is set up portrait (1080x1920) for social. For landscape, change resolution to [1920, 1080] and set the captions position_y to about -380.

Trim Dead Space

Point this at one long recording and it removes the pauses.

The clip is transcribed, then everything between the spoken words is cut and the speech is joined back up. Two settings control how aggressive it is:

  • min_silence_duration (0.8 seconds) is how long a gap has to be before it is cut. Raise it to cut less.
  • padding (0.1 seconds) is how much air is kept around each kept segment. Raise it if words are being clipped.

The input matches a single .mp4 at the top level of the workspace. If you keep several videos there, change it to the file name you want.

Auto-Caption a Clip

One clip in, the same clip out with per-word captions on top. The footage is not re-cut, so this is the one to reach for when the edit is already finished elsewhere.

Caption styling lives on the captions step: font_size, position_y, text_color, highlight_color and outline_color are the ones worth touching first. The full list is in the captions reference, and the Captions Guide covers fixing mishears and repositioning.

To caption only part of the clip, set in and out on the place_clip step.

Next