asciinema-cast-format

Asciinema v3 cast files are NDJSON format with a JSON header line followed by event lines. Parse the first line to extract metadata (version, width, height, timestamp, duration, title, env, theme), then iterate through remaining lines to decode event codes (o for output, i for input, m for metadata, r for resize) with their associated data payloads.

The cast header JSON includes version, width, height, timestamp, duration, title, env (environment variables), and theme. These fields describe the recording environment, dimensions, timing, and playback settings needed to interpret the event stream correctly.

Yes. Cast files work across Linux, macOS, and Windows for documentation parsing, playback analysis, and tooling integration. The structured NDJSON format enables streaming large recordings, extracting specific events, and converting or validating casts without external dependencies.

Asciinema cast events use four codes: 'o' (output—terminal data), 'i' (input—user input), 'm' (metadata—custom key-value pairs), and 'r' (resize—terminal dimension changes). Each event includes a timestamp and payload specific to its code.

Process cast files line-by-line using streaming parsers to avoid loading entire NDJSON into memory. Parse the header once, then iterate through events, filtering or converting only the data you need for analysis, playback, or export workflows.

This Skill documents v3 cast format specifically. The version field in the header identifies the format version, enabling parsers to handle format variations and ensure compatibility across recording and playback tools.