SYSTRAN/faster-whisper
59.6
Adequate · 16 November 2026
3.1k
lines of production code
Python
primary language
5
measurements over time
What this system is
Features
Add Dockerfile and inference script for faster-whisper
A new Dockerfile and infer.py script are added to the docker directory, providing a containerized example for running the faster-whisper library. The setup uses the nvidia/cuda:12.3.2-cudnn9-runtime-ubuntu22.04 base image, installs python3-pip, and installs the faster-whisper package. The included infer.py script demonstrates transcription using the 'tiny' model on a CUDA device with a sample audio file (jfk.flac).
docker · high confidence
Add dedicated benchmarks for WER, speed, and memory usage
The benchmark directory now includes new scripts to evaluate Word Error Rate (using the \jiwer\ library), inference speed, and memory consumption. The WER benchmarks (\wer\_benchmark.py\ and \evaluate\_yt\_commons.py\) assess accuracy on the Librispeech and YouTube-Commons datasets respectively, while \speed\_benchmark.py\ measures execution time and \memory\_benchmark.py\ tracks GPU/CPU resource usage.
benchmark · high confidence
Project initialization and repository structure
The repository is initialized with standard project configuration files, including a .gitignore, CONTRIBUTING.md, MANIFEST.in, and setup.cfg. The setup.py is configured to read the version from the package, set the license to MIT with copyright held by SYSTRAN, require Python 3.9 or greater, and include development dependencies (black, flake8, isort, pytest). The README.md is updated to provide comprehensive installation instructions, benchmark comparisons against other Whisper implementations, and usage examples for both standard and batched transcription.
(repo-wide) · high confidence
Behavioural changes
Assets directory converted to a Python package
The faster\whisper/assets directory now includes an \\init\\_.py file, making it a valid Python package. This change enables the assets to be properly distributed and imported as part of the package structure.
_faster\whisper/assets · high confidence
faster\_whisper v1.2.1: New Tokenizer, VAD, and Audio Decoding Modules
This release introduces dedicated modules for tokenization, voice activity detection (VAD), and audio decoding, alongside a version bump to 1.2.1. The new \faster\_whisper.tokenizer\ module provides a \Tokenizer\ class that wraps the \tokenizers\ library, handling special tokens (like \\<\|transcribe\|\>\, \\<\|startoftranscript\|\>\) and word-splitting logic for various languages. The \faster\_whisper.vad\ module introduces a \VadOptions\ dataclass and \get\_speech\_timestamps\ function, enabling configurable speech segmentation using Silero VAD. The \faster\_whisper.audio\ module is refactored to use PyAV for decoding, adding support for splitting stereo audio channels and improved memory management via garbage collection. Additionally, \faster\_whisper.utils\ exposes \available\_models\ and \download\_model\ functions, mapping model sizes to their Hugging Face repository IDs.
_faster\whisper · high confidence
Test coverage
Initial test suite for transcription, tokenizer, and utility functions
Added a comprehensive test suite covering core library functionality. Tests verify the \available\_models\ and \download\_model\ utilities, tokenizer behavior including suppressed tokens and Unicode splitting, and transcription features such as language detection, batched inference, VAD filtering, stereo channel separation, and multilingual support.
tests · high confidence
Dependencies
Update core dependencies and add benchmark requirements
The project updates several core dependencies: CTranslate2 is upgraded to version 4.x (previously 3.5+), the Hugging Face Hub client is added (\>=0.23), tokenizers is relaxed to allow versions up to 1.x, ONNX Runtime is added (\>=1.14), and PyAV is upgraded to version 11+. Additionally, a new benchmarking environment is introduced via \benchmark/requirements.txt\, specifying tools like \transformers\, \jiwer\, \datasets\, \memory\_profiler\, \py3nvml\, and \pytubefix\ for performance and quality evaluation.
(dependencies) · high confidence
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
Score
- CAI 56 → 60 (+3.2)
- Rubric changed (rubric-2026.11.2 → rubric-2026.11.4) — scores are not directly comparable.
Lenses
- Code Health 89 → 87 (-1.5)
- Architecture 97 → 96 (-0.9)
- Maturity 47 → 55 (+8.4)
- Readiness 52 → 52 (+0.0)
- Security 65 → 65 (+0.0)
Findings
- No change — the same findings as the prior survey.
Changes since last survey
- 4 commits — 3 feature/other, 1 fixes
By area
- (root) — 1 commit
- faster_whisper/feature_extractor.py — 1 commit
- faster_whisper/transcribe.py — 1 commit
- faster_whisper/vad.py — 1 commit
Notable commits
- fix: Fix incorrect float32 dtype check (#1700)
- change: Accept one full window in batched transcription without VAD (#1701)
- change: Make collect_chunks() linear in the number of chunks (#1692)
- change: docs: fix model conversion command (#1696)
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
SYSTRAN/faster-whisper was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 16 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit 6cb639e70786aea42b9233006ae60eaf07190bdb — the exact code this score is about.
- Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
- Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.73+preprod-52de90b74216.