auto_broll_placer

A unified AI agent that reads your full transcript and fills timeline gaps using the best available content type (existing broll clips, web-sourced media, web screenshots, AI-generated video with Veo, AI-generated images with Imagen, or animated drawing overlays), all decided in a single Claude session.

When to use

Use auto_broll_placer as a single step that does the work of broll_place, broll_websource, and drawings_ai_placer combined. Claude sees the whole transcript and timeline state at once and picks the most appropriate content type per gap.

In the editor

Add it from the Library (AI group), or with + add step and a search for auto_broll_placer. It replaces a row of separate b-roll and overlay steps, so a template using it is usually one transcript step, one cut, and this.

Under Inputs, add transcript as a step input. Then set the few parameters that shape the whole thing: max_generate is the total number of clips it may place, min_gap_sec the shortest gap worth filling, and level where everything lands. backbone_step is the ID of the step that sets the video's length.

The most useful control is exclude_tools, which is how you keep it away from the expensive options: excluding the generating tools leaves it to work with footage you already have and free web material. tools is the opposite, restricting it to a named set. Anything whose service you have no key for is left out on its own, so this step still works with no keys set at all.

This step has a lot of parameters, most of them prefixed with the kind of content they apply to and tucked behind Show advanced. You do not need them to start. If you find yourself hunting for one, the YAML tab and its hover help on the processor name lists every setting with its type and default, which is often quicker than scrolling the form.

Tick Review gate. A step that can fill a whole video in one pass is a step you want to see the output of before it becomes the basis for captions and titles.

Everything it places appears on the level you chose in the run timeline, whatever it came from, and you can preview, retime or remove any of it there.

YAML Example

- id: visuals
  processor: auto_broll_placer
  input:
    - id: transcript
      step: generate_transcript
  params:
    level: broll
    backbone_step: ai_cut
    max_generate: 12
    min_gap_sec: 3
    exclude_tools: [broll_veo, broll_imagen]
  output:
    timeline: broll
  cache: true
  review: true

Inputs

Params

Output

EditDecisions: one per placed clip at the configured level. Clips this step generates or downloads are kept for reuse, so re-running it does not pay to fetch them again.

Manifest

A unified manifest tracks the generated/downloaded clips. On re-runs (cache miss), previously generated/downloaded clips are reused, avoiding redundant generation. Ask the AI Assistant to clear the step's saved work, or re-run from that step with the per-step control on the run page, to regenerate everything.

Notes

• Categories whose required service is unavailable (no API key, no renderer) are silently excluded from the tool list, so Claude never sees them.
• tools and exclude_tools are mutually exclusive.
• New tool categories added to the registry in the future are automatically available to auto_broll_placer without any code changes.