distil-whisper

by Hugging Face

Distilled Whisper: 6× faster, 49% smaller, within 1% WER of the teacher.

TL;DR

Distilled Whisper: 6× faster, 49% smaller, within 1% WER of the teacher.

Best for english-only workloads where latency and cost matter more than multilingual coverage. Pricing: free.

Category
Open source
License
MIT
Stars
★ 4.1k
Last push
2025-01-08
Pricing
free
Platforms
Linux, macOS, GPU, CPU

What it is

A Hugging Face distillation of Whisper that keeps most of the accuracy while cutting inference cost. Best when you know your audio is English and you want to serve many concurrent requests on modest hardware. MIT-licensed.

Best for: English-only workloads where latency and cost matter more than multilingual coverage.
Watch out for: English-only in v3; slightly lower WER than Whisper-large-v3 on long-form.

Install / use

pip install transformers  # then load distil-whisper/distil-large-v3

Features

Speaker diarizationNo
Word-level timestampsYes
Streaming / real-timeNo
Languages supported1
HIPAA eligibleNo

Links

GitHub repo ↗

distil-whisper vs Whipscribe

Featuredistil-whisperWhipscribe
CategoryOpen sourceTranscription APIs
Pricingfreefree beta
Speaker diarizationNoYes
Word timestampsYesYes
StreamingNoNo
Languages199
PlatformsLinux, macOS, GPU, CPUWeb, API, MCP
Cost comparison: distil-whisper vs Whipscribe
distil-whisper at free · Whipscribe free beta · 99 languages · data checked 2026-04-24.
How Whipscribe handles your audio
Paste a URL, upload a file, or record — get a diarized transcript in under a minute.

Alternatives to distil-whisper

Whipscribe is a managed faster-whisper + whisperX service. If you want transcripts without running infrastructure, we're one click away.

Try Whipscribe →