assemble_transcript

Merges per-clip Whisper transcripts (from analyze_footage) with backbone-level edit decisions to produce a single timeline-relative transcript with word-level timestamps.

When to use

Use after footage_assembler when you need a unified transcript for captions, broll placement, or video_director. This is the auto-assembly equivalent of generate_transcript.

In the editor

Add it from the Library (Transcript group), or with + add step and a search for assemble_transcript, after the steps that analyzed and assembled your footage.

Under Inputs, add two entries, both of type step: analysis, pointing at your analyze_footage step, and footage, pointing at your footage_assembler step. The one parameter, backbone_level, names the level whose cut the transcript should follow, and main is almost always right.

It puts nothing on the timeline. What it produces is a transcript of the edit as it was actually assembled, which is what a captions step or a b-roll step wants to read. In the run view it reports done, and you then wire it into those steps as their transcript input.

YAML Example

- id: transcript
  processor: assemble_transcript
  input:
    - id: analysis
      step: analyze_raw
    - id: footage
      step: assemble
  params:
    backbone_level: main
  cache: true

Inputs

Params

Output

Analysis containing a merged Transcript with timeline-adjusted word timestamps.

Notes

• No API key required: this is a pure data transformation.
• The output is in the same format as generate_transcript and can be used by all the same downstream processors.