How to run Whisper large-v3-turbo locally

OpenAI · Speech → text · 810M params · MIT (commercial OK)

A pruned 4-decoder-layer Whisper that transcribes ~8x faster than large-v3 with minimal accuracy loss, now the default for real-time work.

What it costs to run — $0 markup

On your own machine — $0. Runs free if you have about 6.0 GB of VRAM (fp16 ~6GB; int8 ~3GB; far faster than large-v3), via faster-whisper / Transformers.
Rent a GPU — from $0.26/hr. Fits RTX 3090 24GB at the direct vendor price ($0 markup) — pay only for the minutes you generate.
Download the weights — free. Open weights at openai/whisper-large-v3-turbo.

Note: generative-media models are billed per image / per second / per minute on hosted services — not per token. Running locally or on your own rented GPU is usually far cheaper and keeps your data on your machine.

Key facts

Does	Speech → text
VRAM to run	~6.0 GB (fp16 ~6GB; int8 ~3GB; far faster than large-v3)
Download	~1.6 GB
Parameters	810M
License	MIT (commercial use OK)
Run with	faster-whisper / Transformers

Get Whisper large-v3-turbo on Hugging Face →

More voice & audio models

Whisper large-v3 — Speech → text, ~10 GB VRAM
Kokoro-82M — Text → speech, ~1.0 GB VRAM
XTTS-v2 — Voice cloning, ~6.0 GB VRAM
MusicGen Large — Text → music, ~16 GB VRAM
Chatterbox — Text → speech, ~8.0 GB VRAM
Parakeet TDT 0.6B v2 — Speech → text, ~4.0 GB VRAM

Browse: all media models · chat / LLM models