broll_screenshot_websource

Generates animated background clips from web screenshots. It takes screenshots of relevant web pages, then renders them as video clips with motion effects (zoom, blur, pixelate) and an optional source attribution card overlay.

When to use

Use when you want web-sourced imagery (article screenshots, product pages, documentation) as animated broll. By default (auto_find: true) the processor uses AI web search to find and screenshot relevant pages based on the transcript context.

In the editor

Add it from the Library (B-Roll group), or with + add step and a search for broll_screenshot_websource.

Under Inputs, add transcript as a step input pointing at your transcript step, not at the step that placed the footage. With auto_find left on, that transcript is what it searches from: it finds pages related to what you are talking about and photographs them. max_clips is the budget, min_gap_sec decides which gaps are worth filling, and min_duration_sec with max_duration_sec bound how long each shot is held.

The look is set by animation_style, zoom_factor and blur_type: a slow push with a little blur reads as a background, a sharp still reads as a screenshot you are asking people to read. source_label and the card settings put a small attribution card on screen. dont_use_domains keeps named sites out of your video, which is worth setting before the first run rather than after seeing a competitor's page in your cut.

The clips appear on the b-roll level in the run timeline. Only public pages can be photographed; anything behind a login is skipped and the step moves on to the next candidate, so a gap left empty here usually means it could not get in.

YAML Example

- id: web_broll
  processor: broll_screenshot_websource
  input:
    - id: transcript
      step: transcribe_main
  params:
    backbone_step: ai_cut
    min_gap_sec: 5
    blur_type: blur
    animation_style: zoom_blur
  output:
    timeline: broll
  cache: true

Inputs

Params

Output

EditDecisions on the configured output_level.

Notes

• Uses Puffin Ship's built-in AI for the agent loop and screenshot content validation (no API key required).
• auto_find: true (default) uses the transcript to search for and screenshot relevant pages. Set auto_find: false and provide URLs via the transcript or prompt to control which pages are used.
• zoom_factor is fractional (0.08 = 8% zoom), not a multiplier: 1.05 would mean 105% zoom, which is excessive.
• Only public web pages can be screenshotted, over http or https on the standard web ports. A URL pointing at a private or internal address is skipped and the step moves on to the next candidate. Use dont_use_domains to exclude public sites you don't want appearing in your video.