Using Cockatiel

Cockatiel turns a recording into a set of time-aligned, speaker-labelled segments you can export to the annotation tool of your choice. Everything runs locally in your browser — your media is never uploaded.

Open Cockatiel and follow along.

Loading media

There are three ways to get a recording into Cockatiel:

• Drop a file — drag an audio or video file onto the drop zone, or click to browse. Video files are supported; Cockatiel uses the audio track for segmentation and shows the video for reference.
• Paste a URL — use the URL input below the drop zone and click Load URL to pull audio from a remote https location.
• Deep link — open Cockatiel with an ?audio= query parameter, for example https://cockatiel.crate-works.org/app/?audio=https://catalog.example.org/path/to/audio.wav. Cockatiel fetches the file, shows its size to confirm, then segments it. The link stays in the address bar so it can be shared.

Sessions sourced from a URL are stored against that URL — revisiting the same link restores your saved transcript instantly and re-downloads the audio in the background.

Loading from a catalogue? The remote server must allow cross-origin reads from Cockatiel’s origin (Access-Control-Allow-Origin, plus GET and a Range header — Cockatiel issues a small range request to discover the file size). Only https: URLs are accepted.

Automatic segmentation

When you load a recording, Cockatiel runs voice-activity detection to split it into utterances:

• Silero VAD (WebAssembly) is the default — accurate neural speech detection that runs entirely in the browser.
• An energy-based fallback is used where the WASM model can’t run.

You can re-run segmentation with different sensitivity if the defaults over- or under-split.

Editing on the waveform

Every segment is a region on the waveform. Work directly on it:

• Drag a region to move it, or drag its edges to resize.
• Split a region at the playhead.
• Merge adjacent regions into one.
• Delete a region you don’t need.

The audio playback and your edits stay in sync, so you can fine-tune boundaries by ear.

Speakers

Assign each segment to a speaker. Speakers have configurable names and colours, so a multi-speaker recording stays readable at a glance and the speaker labels carry through to your export.

Looping playback

Click a region to loop it while you transcribe. The segment repeats so you can keep your hands on the keyboard and type what you hear without constantly seeking.

Keyboard shortcuts

Cockatiel is built for a keyboard-driven workflow:

Press ? at any time to see the complete, up-to-date list.

Exporting

When your transcript is ready, export it to the format your analysis tools expect:

• EAF — ELAN annotation format
• SRT — subtitles
• TextGrid — Praat
• CSV — spreadsheets / scripting
• Plain text

Exports include your segment boundaries and speaker labels, ready to open in the destination tool.

Catalogue integration

If your deployment is configured with one or more catalogue providers, you can browse and load items directly from an RO-Crate catalogue (such as PARADISEC or LDaCA), and import existing ELAN transcripts attached to an item. See Self-host for how to point Cockatiel at your own catalogue.