Parse

The .parse() stage reads a Playwright trace archive and extracts all the data needed for video generation.

Basic usage

import { Recast } from 'playwright-recast'

const pipeline = Recast
  .from('./test-results/trace.zip')
  .parse()

You can also point to a directory containing trace files:

const pipeline = Recast
  .from('./test-results/')
  .parse()

The parser reads the trace.zip and produces a structured representation containing:

Data	Source	Used by
Actions	User interactions (click, fill, type, navigation)	Speed processing, click effects, auto-zoom
Screenshots	Frame-by-frame page captures	Video frames
Network events	API calls, resource loading	Speed classification (network wait)
Cursor positions	Mouse coordinates at each action	Cursor overlay, click effects, auto-zoom
BDD step titles	playwright-bdd Gherkin step names	Subtitle generation via `.subtitlesFromTrace()`

# Parse is implicit in CLI mode — it always runs
npx playwright-recast -i ./test-results/trace.zip -o demo.mp4

The renderer needs a source .mp4/.webm to apply speed, zoom, overlays, and subtitles against. Two layouts are supported:

Sibling .webm (preferred) — when your Playwright config enables recordVideo (e.g. use: { video: 'on' }), the trace's sibling .webm is auto-discovered. Native frame rate, best quality.
Standalone trace.zip — when no sibling .webm exists (trace downloaded from the trace viewer, or test run without recordVideo), the pipeline assembles a CFR 25fps .mp4 from the screencast JPEG frames stored inside the trace. Variable capture cadence — .webm is preferred when available.

The trace file must be generated with Playwright's tracing enabled (--trace on or via playwright.config.ts).
.parse() should always be the first stage after .from(). Other stages depend on the parsed trace data.
Parsing is fast — the trace is read once and the structured data is passed through the pipeline.