How kb-labb/easytranscriber works
easytranscriber turns audio and optional metadata into transcriptions with word-level timestamps, using configurable ASR backends and GPU-accelerated forced alignment. Its built-in easysearch indexes transcription chunks or alignments in SQLite and provides a browser interface for querying transcripts and playing audio. The pipeline and search entry-point excerpts are partial; the README documents alignment and search capabilities, while some internal wiring is not shown, so those relationships are omitted when unsupported.
Main entry points: pipelines.py Pipeline · app.py Search web app · datamodel.py Transcript outputs · ct2.py CTranslate2 ASR
Components 21
Transcription pipeline
- Pipeline
src/easytranscriber/pipelines.py - Voice detection
src/easytranscriber/pipelines.py - CTranslate2 ASR
src/easytranscriber/asr/ct2.py - Hugging Face ASR
src/easytranscriber/asr/hf.py - Cohere ASR
src/easytranscriber/asr/cohere.py
Audio and data
- Audio reader
src/easytranscriber/audio.py - Audio datasets
src/easytranscriber/data/dataset.py - Batch collators
src/easytranscriber/data/collators.py - Speech data model
src/easytranscriber/data/datamodel.py
Transcript search
- easysearch command
src/easytranscriber/search/__main__.py - Transcript indexer
src/easytranscriber/search/indexer.py - SQLite search index
src/easytranscriber/search/db.py - Search web app
src/easytranscriber/search/app.py - Search interface
src/easytranscriber/search/templates
Alignment and text
- Text normalization
src/easytranscriber/text/normalization.py - Forced alignment
src/easytranscriber/pipelines.py - Transcript outputs
src/easytranscriber/data/datamodel.py
Other components
- Audio user
- Audio files
- Search user
- Audio playback files
Connections 25
- Audio user → Pipeline: submits audio
README.md - Audio files → Pipeline: provides input
README.md - Pipeline → Voice detection: runs detection
src/easytranscriber/pipelines.py - Pipeline → Audio datasets: loads chunks
src/easytranscriber/pipelines.py - Audio datasets → Audio reader: reads segments
src/easytranscriber/data/dataset.py - Audio datasets → Speech data model: uses metadata
src/easytranscriber/data/dataset.py - Pipeline → CTranslate2 ASR: dispatches ASR
src/easytranscriber/pipelines.py - Pipeline → Hugging Face ASR: dispatches ASR
src/easytranscriber/pipelines.py - Pipeline → Cohere ASR: dispatches ASR
src/easytranscriber/pipelines.py - CTranslate2 ASR → Batch collators: batches features
src/easytranscriber/asr/ct2.py - Hugging Face ASR → Batch collators: batches features
src/easytranscriber/asr/hf.py - Cohere ASR → Batch collators: batches audio
src/easytranscriber/asr/cohere.py - Pipeline → Text normalization: configures normalization
src/easytranscriber/pipelines.py - Pipeline → Forced alignment: performs alignment
README.md - Forced alignment → Transcript outputs: produces timestamps
README.md - CTranslate2 ASR → Transcript outputs: saves transcription
src/easytranscriber/asr/ct2.py - Hugging Face ASR → Transcript outputs: saves transcription
src/easytranscriber/asr/hf.py - Cohere ASR → Transcript outputs: saves transcription
src/easytranscriber/asr/cohere.py - Search user → easysearch command: starts search
README.md - easysearch command → Transcript indexer: indexes files
src/easytranscriber/search/__main__.py - Transcript indexer → SQLite search index: writes index
src/easytranscriber/search/indexer.py - easysearch command → Search web app: starts server
src/easytranscriber/search/__main__.py - Search web app → SQLite search index: queries index
src/easytranscriber/search/app.py - Search web app → Search interface: renders pages
src/easytranscriber/search/app.py - Search web app → Audio playback files: serves audio
src/easytranscriber/search/app.py