OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Audio & Voice AI

Distil Whisper

Distil-Whisper is an open-source, distilled version of Whisper for English speech recognition. Its Python repository provides checkpoints and Hugging Face Transformers examples for transcribing audio, including short clips and local files.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Distil Whisper — OpenViST project profile in Audio & Voice AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Distil-Whisper when building an English transcription workflow with Transformers, especially when you want to compare smaller distilled checkpoints or process clips through a Python pipeline.

Who it is for

  • Developers adding English speech-to-text to Python applications.
  • Teams comparing distilled Whisper checkpoints for transcription workflows.
  • Builders considering a smaller checkpoint for resource-constrained applications.

Documented uses

  • Turn English audio samples or local audio files into text.
  • Choose among distilled checkpoints based on model size and reported transcription results.
  • Integrate a speech-recognition model through Hugging Face Transformers.
Before you choose

Requirements & limits

The repository is public and lists the MIT license.

  • The model is limited to English speech recognition; the README points to Whisper Turbo for multilingual recognition.
  • The repository recommends distil-large-v3 for most applications and identifies the 166-million-parameter distil-small.en checkpoint for resource-constrained use.
  • The documented Transformers usage requires version 4.35 or later.
  • The short-form example covers audio under 30 seconds; the repository also discusses long-form evaluation.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Transcribe a short audio clip with Transformers

Use the documented automatic speech-recognition pipeline to transcribe an audio sample or local file under 30 seconds.

Before you start

  • Python environment with PyTorch and Hugging Face Transformers 4.35 or later.
  • Install Transformers, Accelerate and Datasets with audio support as shown in the repository; the example uses CUDA when available and otherwise CPU.
  • An audio sample or local audio file.

Steps

  1. Install or upgrade the listed packages: transformers, accelerate, and datasets[audio].

  2. Load distil-whisper/distil-large-v3 with AutoModelForSpeechSeq2Seq and AutoProcessor, selecting the device and tensor precision as in the README.

  3. Create an automatic-speech-recognition pipeline using the loaded model and processor tokenizer and feature extractor.

  4. Pass an audio sample or local audio path to the pipeline and read the returned text.

Expected outcome

A transcription text result for the supplied English audio clip.

Guide limitations

  • The documented short-form path is for clips under 30 seconds.
  • Distil-Whisper supports English speech recognition only.
  • The repository’s example uses local Python code; it does not document a hosted transcription service.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
huggingface/distil-whisper · Official homepage title · retained in its source language
How we review sourcesSuggest a correction