faster-whisper-large-v3

faster-whisper-large-v3

Turns speech in audio files into timestamped text, as OpenAI's Whisper large-v3 converted for CTranslate2 and faster-whisper. For developers who transcribe recordings in many languages on their own machine.

Made by Systran on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: This repository holds OpenAI's Whisper large-v3 speech recognition model, converted to the CTranslate2 format. You can use it in CTranslate2 or in projects built on it, such as faster-whisper. It transcribes audio to text with start and end times for each segment.

Who it suits: Developers, creators and researchers who want to transcribe audio locally and can run a Python script.

Industry | Who uses it | How they use it | Example
Media | Podcasters and video editors | Transcribe recordings into text | Turn an interview into captions
Research | Interviewers and analysts | Convert spoken material into searchable text | Transcribe field recordings
Software | Developers | Add speech to text to an app | Build a transcription feature around faster-whisper

If you only need a hosted transcription service with no setup, you can skip it.

Where to use it:

• Your own computer with Python, on Windows, macOS or Linux
• CTranslate2, the runtime this format is made for
• faster-whisper, the project the card names

Use case | Who | What they get back
Transcription | Creators | Text for each audio segment
Captions | Video editors | Segment start and end times with text
Multilingual audio | Researchers | Transcripts across many languages

How to set it up:

• Python with pip
• The faster-whisper package, which provides WhisperModel
• An audio file such as an mp3

1. Install the Hugging Face hub tool.
2. Download the model files, or let faster-whisper fetch them by name.
3. Run the transcription script below.

pip install -U huggingface_hub
hf download Systran/faster-whisper-large-v3

How to use it: Load WhisperModel with the large-v3 name, call transcribe on an audio file, then loop over the segments and print each one with its times.

from faster_whisper import WhisperModel

model = WhisperModel("large-v3")

segments, info = model.transcribe("audio.mp3")
for segment in segments:
    print("[%.2fs -> %.2fs] %s" % (segment.start, segment.end, segment.text))

Example: You run the script on a file named audio.mp3. It prints one line per segment, each with a start time, an end time and the spoken text.

For newcomers and experts: If you are new, run the script on a short clip first. The first run needs the model files, which are large.

For experts, the conversion used ct2-transformers-converter with float16 quantization, copying the tokenizer and preprocessor config. Weights are saved in FP16, and you can change that when loading through the compute_type option in CTranslate2.

OpenAI Whisper large-v3 is the source
ct2-transformers-converter converts it
CTranslate2 or faster-whisper loads it
You get timestamped text

Good to know: The card shows the conversion command it used.

ct2-transformers-converter --model openai/whisper-large-v3 --output_dir faster-whisper-large-v3 \
    --copy_files tokenizer.json preprocessor_config.json --quantization float16

• License: MIT, as the Hugging Face record lists it.
• Cost to run: no API key or account is named. You need a computer that can run the model.
• Data it reaches: it reads the audio file you give it. Downloading reaches Hugging Face.
• Not for: the card sends you to the original Whisper large-v3 card for details about the model itself.

On Hugging Face, 13 November 2026

• Task: automatic speech recognition
• Library: ctranslate2
• Downloads in the last 30 days: 1,660,044
• Likes: 714
• License: mit
• Languages: en, zh, de, es, ru, ko, fr, ja
• Fine-tunes: 1
• Tags: ctranslate2, audio, automatic-speech-recognition

Links

• Model: https://huggingface.co/Systran/faster-whisper-large-v3
• Files: https://huggingface.co/Systran/faster-whisper-large-v3/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download Systran/faster-whisper-large-v3

Then follow the usage in its model card. Nothing in this kit runs by itself.