Status: Current Last updated: 2026-09-07
This repository contains Batchalign: the audio and ML pipeline for producing and enriching CHAT transcripts. It supports transcription, forced alignment, morphotagging, utterance segmentation, translation, diarization, comparison, and the Batchalign desktop shell.
On macOS or Linux:
curl -LsSf https://raw.githubusercontent.com/TalkBank/batchalign/main/bootstrap/bootstrap.sh | shOn Windows, from PowerShell:
powershell -NoProfile -ExecutionPolicy Bypass -Command "Invoke-Expression ((Invoke-WebRequest -UseBasicParsing 'https://raw.githubusercontent.com/TalkBank/batchalign/main/bootstrap/bootstrap.ps1').Content)"The bootstrapper installs uv
when needed, then installs or upgrades batchalign[all] on Python 3.11 with
prerelease versions enabled. The installer requires Batchalign 0.10 or newer.
Next, follow the Quick Start. If anything is confusing, see the Batchalign documentation.
The CHAT parser, typed model, transformations, and grammar come from the version-pinned TalkBank Chatter repository. They are dependencies of Batchalign rather than separately shipped products in this repository.
Use the Bazel-backed just recipes:
just batchalign build debug
just batchalign test debug
just batchalign pytest
just batchalign cli -- --helpjust batchalign cli is the authoritative development entrypoint for the CLI.
Bazel uses host-adaptive parallelism by default. Set BATCHALIGN_BAZEL_JOBS
explicitly when a resource-constrained runner needs a lower limit.
The main source areas are:
crates/batchalign/— typed Rust core and PyO3 enginepython/batchalign/— Python backends and CLIapps/batchalign/— desktop applicationresources/test_fixtures/— integration and parity fixturesscripts/parity/— bounded pre/post parity tooling
The website is an mdBook. Edit the homepage in book/src/index.md, the other
pages in book/src/, and navigation in book/src/SUMMARY.md.
just docs setup # once: install mdBook and its plugins; requires Rust/Cargo
just docs serve # http://localhost:3000/; automatically reloads edits
# Or choose another port: just docs serve 8000just docs build builds the website into book/build/html/.
Live previews use separate book/build/preview-<port>/ directories, so builds
and previews can run at the same time.
The docs workflow checks pull requests and deploys that directory on pushes to
main. Enable Settings → Pages → Build and deployment → Source: GitHub
Actions once in the repository before deploying.
The homepage loads book/src/downloads.js and shows macOS, Windows, and Linux
download buttons. It scans published stable releases, falling back to older
releases when necessary. macOS links to a DMG, Windows to an x86-64 installer,
and Linux offers only x86-64 DEB and RPM packages. Other builds are available
through the all-releases link. Each main button shows its installer version.
Download URLs come from assets[].browser_download_url, so links update
without rebuilding the site. A manual releases link works when JavaScript or
the API is unavailable. To inspect the latest tag yourself:
curl -fsSL https://api.github.com/repos/TalkBank/batchalign/releases/latest | jq -r .tag_nameBSD-3-Clause. Copyright (c) 2026, Carnegie Mellon University.