ASR-Kit

Modular Python library for Automatic Speech Recognition. Pass WAV(s) to a unified Transcriber; get TranscriptionResult(s) back. Backed by pluggable model drivers.

Supported: Qwen3-ASR | Whisper (planned) | Parakeet (planned)

Setup

# 1. Create and activate a dedicated conda environment (Python 3.12 recommended)
conda create -n asr-kit python=3.12
conda activate asr-kit

# 2. Install directly from GitHub
pip install "asr-kit[qwen] @ git+https://github.com/dqvid3/asr-kit.git"

# 3. (Optional - Qwen only) Install Flash Attention 2 for Ampere+ GPUs (RTX 3090/4090/ADA/A100)
# Significantly reduces VRAM and speeds up long audio processing when using the Qwen model.
# If your machine has less than 96GB of RAM and lots of CPU cores, run:
# MAX_JOBS=4 pip install -U flash-attn --no-build-isolation
pip install -U flash-attn --no-build-isolation

# To remove the environment later
conda env remove -n asr-kit

Note: Python 3.13 may have compatibility issues with torch or qwen-asr. Use 3.12 to be safe.

Usage

from asr_kit import Transcriber

t = Transcriber(model="qwen", device="cuda")

# Single file → TranscriptionResult
result = t.transcribe("audio.wav")
print(result.text)

# Multiple files → list[TranscriptionResult]
results = t.transcribe(["a.wav", "b.wav"], language="English")

# With word-level timestamps (Qwen: requires use_forced_aligner=True)
t = Transcriber(model="qwen", device="cuda", use_forced_aligner=True)
result = t.transcribe("audio.wav", return_timestamps=True)
for word in result.timestamps:
    print(f"{word.start:.2f}s  {word.text}")

Output: TranscriptionResult

field	type	description
text	str	Full transcript
language	str | None	Detected language code
timestamps	list[WordTimestamp] | None	Word-level timing (if requested)
audio_path	str	Absolute source WAV path
model	str	Model identifier

WordTimestamp: text: str, start: float, end: float (seconds)

Adding a Driver

1. Create src/asr_kit/drivers/my_driver.py implementing BaseDriver
2. Implement load_model(**kwargs) -> None and transcribe(audio_paths, **kwargs) -> list[TranscriptionResult]
3. Register the key in _DRIVER_REGISTRY dict in transcriber.py
4. Add optional deps under [project.optional-dependencies] in pyproject.toml

Try It

After cloning the repo:

python examples/qwen_basic.py audio.wav
python examples/qwen_basic.py audio.wav --timestamps
python examples/qwen_basic.py a.wav b.wav --language English

Issues and PRs welcome.

License

MIT