Whisper v3 (local)
Whisper is a general-purpose speech recognition model that performs multilingual speech recognition, speech translation, and language identification. It is trained on a large dataset of diverse audio and uses a Transformer sequence-to-sequence architecture. It is for developers who need robust speech-to-text capabilities in their applications.
🎬 How to Install & Use Whisper AI Voice to Text · Kevin Stratvert
✨ Key features
- Multilingual speech recognition and translation
- Language identification and voice activity detection
- Multiple model sizes for speed/accuracy tradeoffs
- Command-line interface and Python API
- Handles various audio formats via ffmpeg
- Open-source with MIT license
🎯 Use cases
- Transcribing audio files into text
- Translating non-English speech into English
- Building voice-controlled applications
- Creating subtitles or captions for videos
- Analyzing spoken language in audio data
📦 Installation
🧰 Requirements: Python 3.8-3.11, PyTorch, ffmpeg, and optionally Rust if tiktoken wheel is not available.
pip install -U openai-whisper
Alternatively, install from GitHub:
pip install git+https://github.com/openai/whisper.git
To update to the latest commit:
pip install --upgrade --no-deps --force-reinstall git+https://github.com/openai/whisper.git
Also install ffmpeg (example for Ubuntu):
sudo apt update && sudo apt install ffmpeg
If tiktoken fails, install Rust and setuptools-rust.
🚀 Usage
whisper audio.flac audio.mp3 audio.wav --model turbo
Python usage:
import whisper
model = whisper.load_model("turbo")
result = model.transcribe("audio.mp3")
print(result["text"])
⚠️ Good to know
The turbo model is not trained for translation tasks; use multilingual models for translation. Performance varies by language.
❓ FAQ
What is the difference between English-only and multilingual models?
English-only models (e.g., tiny.en) are optimized for English and perform better for English-only tasks. Multilingual models support multiple languages and can also translate to English.
How do I translate non-English speech to English?
Use a multilingual model (e.g., medium or large) with the --task translate option, as the turbo model does not support translation.
What are the system requirements?
Python 3.8-3.11, PyTorch, ffmpeg, and optionally Rust if tiktoken wheel is not available. VRAM requirements vary by model size, from ~1 GB for tiny to ~10 GB for large.
Can I use Whisper in Python?
Yes, you can load a model and transcribe audio using the Python API, as shown in the README.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.