Do I need a GPU to run Whisper?

Whisper can run on a CPU, but it is 10-20x faster on a GPU. For the 'large' and 'turbo' models, a GPU with at least 6-10GB of VRAM is highly recommended.

How accurate is the translation feature?

Whisper is highly effective at translating speech from many languages into English, as it was trained on 680,000 hours of multilingual and multitask supervised data.

Is there a limit to the length of audio I can transcribe?

Whisper can handle long audio files, but for optimal accuracy and memory management, it is often best to process files in 30-minute segments or use the provided batch processing patterns.

Can I use Whisper for real-time transcription?

While OpenAI's standard Whisper is optimized for batch processing, this skill also provides guidance on using faster-whisper for low-latency and streaming applications.

What audio formats does the Whisper skill support?

Whisper supports most common audio and video formats, including MP3, MP4, WAV, and AAC, provided that ffmpeg is installed on your system.

Whisper Speech Recognition

Name: Whisper Speech Recognition
Author: Orchestra-Research

byOrchestra-Research

•

3,983

•

数据科学与机器学习

Transcribes and translates audio across 99 languages using OpenAI's robust general-purpose speech recognition models.

Whisper is a high-performance Claude Code skill designed for developers and researchers who need reliable speech-to-text capabilities. It leverages OpenAI's pre-trained models to provide accurate transcription, automatic language identification, and seamless translation of foreign audio into English. With six model sizes ranging from the resource-efficient 'tiny' to the high-fidelity 'large' and 'turbo' versions, this skill enables a wide variety of audio processing workflows including podcast transcription, meeting automation, and multilingual data collection, even in noisy environments.

主要功能

01Multilingual support for 99 languages with automatic language identification.

02Flexible model scaling from 39M to 1550M parameters to balance speed and accuracy.

03Robust performance against background noise and various accents through large-scale training.

04English translation task for converting foreign speech directly into English text.

05Generates detailed timestamps and word-level alignment for subtitle creation.

063,983 GitHub stars

使用场景

01Building research pipelines that process and analyze multilingual audio datasets at scale.

02Extracting structured meeting notes and action items from recorded voice conversations.

03Automating the generation of subtitles and searchable transcripts for video and podcast content.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add orchestra-research/ai-research-skills whisper

For use in Claude.ai and ChatGPT

Download Skill