Whisper

orchestra-research/ai-research-skills/18-multimodal/whisper

by orchestra-research773a52944ba4MIT13K starsListed Oct 8, 2026Updated Oct 8, 2026Repository updated 3 months ago

OpenAI's general-purpose speech recognition model. Supports 99 languages, transcription, translation to English, and language identification. Six model sizes from tiny (39M params) to large (1550M params). Use for speech-to-text, podcast transcription, or multilingual audio processing. Best for robust, multilingual ASR.

Instructions onlyAI & Agents

AI-generated overview

Guides use of OpenAI's Whisper speech recognition model for transcription, translation to English, and language identification.

What it does: This skill documents how to run OpenAI's Whisper speech recognition model for speech-to-text, translation to English, and language identification across 99 languages. It covers model sizes, transcription options such as language selection, initial prompts, word-level timestamps, and temperature fallback, plus command-line usage and output formats like txt, srt, vtt, and json. It also describes batch processing, GPU acceleration, subtitle generation, and integration with tools such as LangChain.
When to use it: Use it for speech-to-text transcription, podcast or video transcription, meeting notes automation, translation of audio to English, or multilingual and noisy audio processing. It is intended when robust multilingual ASR is needed rather than real-time streaming or speaker diarization.
Requirements: Requires Python 3.8-3.11, the openai-whisper package, and ffmpeg; the document also mentions transformers and torch as dependencies and faster-whisper for streaming. GPU acceleration is optional. It ships no scripts, only instructions and a language reference file.

Whisper - Robust Speech Recognition

OpenAI's multilingual speech recognition model.

When to use Whisper

Use when:

Speech-to-text transcription (99 languages)
Podcast/video transcription
Meeting notes automation
Translation to English
Noisy audio transcription
Multilingual audio processing

Metrics:

72,900+ GitHub stars
99 languages supported
Trained on 680,000 hours of audio
MIT License

Use alternatives instead:

AssemblyAI: Managed API, speaker diarization
Deepgram: Real-time streaming ASR
Google Speech-to-Text: Cloud-based

Quick start

Installation

bash

# Requires Python 3.8-3.11pip install -U openai-whisper
# Requires ffmpeg# macOS: brew install ffmpeg# Ubuntu: sudo apt install ffmpeg# Windows: choco install ffmpeg

# Requires Python 3.8-3.11pip install -U openai-whisper
# Requires ffmpeg# macOS: brew install ffmpeg# Ubuntu: sudo apt install ffmpeg# Windows: choco install ffmpeg

Basic transcription

python

import whisper
# Load modelmodel = whisper.load_model("base")
# Transcriberesult = model.transcribe("audio.mp3")
# Print textprint(result["text"])
# Access segmentsfor segment in result["segments"]:    print(f"[{segment['start']:.2f}s - {segment['end']:.2f}s] {segment['text']}")

import whisper
# Load modelmodel = whisper.load_model("base")
# Transcriberesult = model.transcribe("audio.mp3")
# Print textprint(result["text"])
# Access segmentsfor segment in result["segments"]:    print(f"[{segment['start']:.2f}s - {segment['end']:.2f}s] {segment['text']}")

Model sizes

python

# Available modelsmodels = ["tiny", "base", "small", "medium", "large", "turbo"]
# Load specific modelmodel = whisper.load_model("turbo")  # Fastest, good quality

# Available modelsmodels = ["tiny", "base", "small", "medium", "large", "turbo"]
# Load specific modelmodel = whisper.load_model("turbo")  # Fastest, good quality

Recommendation: Use turbo for best speed/quality, base for prototyping

Transcription options

Language specification

python

# Auto-detect languageresult = model.transcribe("audio.mp3")
# Specify language (faster)result = model.transcribe("audio.mp3", language="en")
# Supported: en, es, fr, de, it, pt, ru, ja, ko, zh, and 89 more

# Auto-detect languageresult = model.transcribe("audio.mp3")
# Specify language (faster)result = model.transcribe("audio.mp3", language="en")
# Supported: en, es, fr, de, it, pt, ru, ja, ko, zh, and 89 more

Task selection

python

# Transcription (default)result = model.transcribe("audio.mp3", task="transcribe")
# Translation to Englishresult = model.transcribe("spanish.mp3", task="translate")# Input: Spanish audio → Output: English text

# Transcription (default)result = model.transcribe("audio.mp3", task="transcribe")
# Translation to Englishresult = model.transcribe("spanish.mp3", task="translate")# Input: Spanish audio → Output: English text

Initial prompt

python

# Improve accuracy with contextresult = model.transcribe(    "audio.mp3",    initial_prompt="This is a technical podcast about machine learning and AI.")
# Helps with:# - Technical terms# - Proper nouns# - Domain-specific vocabulary

# Improve accuracy with contextresult = model.transcribe(    "audio.mp3",    initial_prompt="This is a technical podcast about machine learning and AI.")
# Helps with:# - Technical terms# - Proper nouns# - Domain-specific vocabulary

Timestamps

python

# Word-level timestampsresult = model.transcribe("audio.mp3", word_timestamps=True)
for segment in result["segments"]:    for word in segment["words"]:        print(f"{word['word']} ({word['start']:.2f}s - {word['end']:.2f}s)")

# Word-level timestampsresult = model.transcribe("audio.mp3", word_timestamps=True)
for segment in result["segments"]:    for word in segment["words"]:        print(f"{word['word']} ({word['start']:.2f}s - {word['end']:.2f}s)")

Temperature fallback

python

# Retry with different temperatures if confidence lowresult = model.transcribe(    "audio.mp3",    temperature=(0.0, 0.2, 0.4, 0.6, 0.8, 1.0))

# Retry with different temperatures if confidence lowresult = model.transcribe(    "audio.mp3",    temperature=(0.0, 0.2, 0.4, 0.6, 0.8, 1.0))

Command line usage

bash

# Basic transcriptionwhisper audio.mp3
# Specify modelwhisper audio.mp3 --model turbo
# Output formatswhisper audio.mp3 --output_format txt     # Plain textwhisper audio.mp3 --output_format srt     # Subtitleswhisper audio.mp3 --output_format vtt     # WebVTTwhisper audio.mp3 --output_format json    # JSON with timestamps
# Languagewhisper audio.mp3 --language Spanish
# Translationwhisper spanish.mp3 --task translate

# Basic transcriptionwhisper audio.mp3
# Specify modelwhisper audio.mp3 --model turbo
# Output formatswhisper audio.mp3 --output_format txt     # Plain textwhisper audio.mp3 --output_format srt     # Subtitleswhisper audio.mp3 --output_format vtt     # WebVTTwhisper audio.mp3 --output_format json    # JSON with timestamps
# Languagewhisper audio.mp3 --language Spanish
# Translationwhisper spanish.mp3 --task translate

Batch processing

python

import os
audio_files = ["file1.mp3", "file2.mp3", "file3.mp3"]
for audio_file in audio_files:    print(f"Transcribing {audio_file}...")    result = model.transcribe(audio_file)
    # Save to file    output_file = audio_file.replace(".mp3", ".txt")    with open(output_file, "w") as f:        f.write(result["text"])

import os
audio_files = ["file1.mp3", "file2.mp3", "file3.mp3"]
for audio_file in audio_files:    print(f"Transcribing {audio_file}...")    result = model.transcribe(audio_file)
    # Save to file    output_file = audio_file.replace(".mp3", ".txt")    with open(output_file, "w") as f:        f.write(result["text"])

Real-time transcription

python

# For streaming audio, use faster-whisper# pip install faster-whisper
from faster_whisper import WhisperModel
model = WhisperModel("base", device="cuda", compute_type="float16")
# Transcribe with streamingsegments, info = model.transcribe("audio.mp3", beam_size=5)
for segment in segments:    print(f"[{segment.start:.2f}s -> {segment.end:.2f}s] {segment.text}")

# For streaming audio, use faster-whisper# pip install faster-whisper
from faster_whisper import WhisperModel
model = WhisperModel("base", device="cuda", compute_type="float16")
# Transcribe with streamingsegments, info = model.transcribe("audio.mp3", beam_size=5)
for segment in segments:    print(f"[{segment.start:.2f}s -> {segment.end:.2f}s] {segment.text}")

GPU acceleration

python

import whisper
# Automatically uses GPU if availablemodel = whisper.load_model("turbo")
# Force CPUmodel = whisper.load_model("turbo", device="cpu")
# Force GPUmodel = whisper.load_model("turbo", device="cuda")
# 10-20× faster on GPU

import whisper
# Automatically uses GPU if availablemodel = whisper.load_model("turbo")
# Force CPUmodel = whisper.load_model("turbo", device="cpu")
# Force GPUmodel = whisper.load_model("turbo", device="cuda")
# 10-20× faster on GPU

Integration with other tools

Subtitle generation

bash

# Generate SRT subtitleswhisper video.mp4 --output_format srt --language English
# Output: video.srt

# Generate SRT subtitleswhisper video.mp4 --output_format srt --language English
# Output: video.srt

With LangChain

python

from langchain.document_loaders import WhisperTranscriptionLoader
loader = WhisperTranscriptionLoader(file_path="audio.mp3")docs = loader.load()
# Use transcription in RAGfrom langchain_chroma import Chromafrom langchain_openai import OpenAIEmbeddings
vectorstore = Chroma.from_documents(docs, OpenAIEmbeddings())

from langchain.document_loaders import WhisperTranscriptionLoader
loader = WhisperTranscriptionLoader(file_path="audio.mp3")docs = loader.load()
# Use transcription in RAGfrom langchain_chroma import Chromafrom langchain_openai import OpenAIEmbeddings
vectorstore = Chroma.from_documents(docs, OpenAIEmbeddings())

Extract audio from video

bash

# Use ffmpeg to extract audioffmpeg -i video.mp4 -vn -acodec pcm_s16le audio.wav
# Then transcribewhisper audio.wav

# Use ffmpeg to extract audioffmpeg -i video.mp4 -vn -acodec pcm_s16le audio.wav
# Then transcribewhisper audio.wav

Best practices

Use turbo model - Best speed/quality for English
Specify language - Faster than auto-detect
Add initial prompt - Improves technical terms
Use GPU - 10-20× faster
Batch process - More efficient
Convert to WAV - Better compatibility
Split long audio - <30 min chunks
Check language support - Quality varies by language
Use faster-whisper - 4× faster than openai-whisper
Monitor VRAM - Scale model size to hardware

Performance

Real-time factor: 0.1 = 10× faster than real-time

Language support

Top-supported languages:

English (en)
Spanish (es)
French (fr)
German (de)
Italian (it)
Portuguese (pt)
Russian (ru)
Japanese (ja)
Korean (ko)
Chinese (zh)

Full list: 99 languages total

Limitations

Hallucinations - May repeat or invent text
Long-form accuracy - Degrades on >30 min audio
Speaker identification - No diarization
Accents - Quality varies
Background noise - Can affect accuracy
Real-time latency - Not suitable for live captioning

Resources

GitHub: https://github.com/openai/whisper ⭐ 72,900+
Paper: https://arxiv.org/abs/2212.04356
Model Card: https://github.com/openai/whisper/blob/main/model-card.md
Colab: Available in repo
License: MIT

Source and attribution

Source:orchestra-research/ai-research-skillsin18-multimodal/whisperat commit773a529

License: MIT

Content belongs to its original authors. SourceWeft indexes it from a public repository.

Report or request removal

More from orchestra-research/ai-research-skills

Speculative Decoding

orchestra-research

Guides accelerating LLM inference with speculative decoding, Medusa multi-head prediction, and lookahead decoding.

AI & Agents13Kupdated 3 months ago

Systems Paper Writing

orchestra-research

Guides paragraph-level structure and page budgets for writing systems conference papers such as OSDI, SOSP, ASPLOS and NSDI.

Writing & Content13Kupdated 3 months ago

Moe Training

orchestra-research

Guides training and inference of Mixture of Experts models with DeepSpeed or HuggingFace, covering routing, load balancing and expert parallelism.

AI & Agents13Kupdated 3 months ago

Presenting Conference Talks

orchestra-research

Generates conference presentation slides (Beamer LaTeX PDF and editable PPTX) from a compiled paper with speaker notes and talk script. Use when preparing oral talks, spotlight presentations, or invited talks for ML and systems conferences.

Awaiting classification13Kupdated 3 months ago

Model Pruning

orchestra-research

Guides LLM pruning with Wanda, SparseGPT, magnitude and N:M methods to cut model size and speed inference.

AI & Agents13Kupdated 3 months ago

Outlines

orchestra-research

Guides use of the Outlines library for structured, schema-constrained text generation with local models.

AI & Agents13Kupdated 3 months ago

Model	Parameters	English-only	Multilingual	Speed	VRAM
tiny	39M	✓	✓	~32x	~1 GB
base	74M	✓	✓	~16x	~1 GB
small	244M	✓	✓	~6x	~2 GB
medium	769M	✓	✓	~2x	~5 GB
large	1550M	✗	✓	1x	~10 GB
turbo	809M	✗	✓	~8x	~6 GB

Model	Real-time factor (CPU)	Real-time factor (GPU)
tiny	~0.32	~0.01
base	~0.16	~0.01
turbo	~0.08	~0.01
large	~1.0	~0.05

Whisper

Whisper - Robust Speech Recognition

When to use Whisper

Quick start

Installation

Basic transcription

Model sizes

Transcription options

Language specification

Task selection

Initial prompt

Timestamps

Temperature fallback

Command line usage

Batch processing

Real-time transcription

GPU acceleration

Integration with other tools

Subtitle generation

With LangChain

Extract audio from video

Best practices

Performance

Language support

Limitations

Resources

Source and attribution

More from orchestra-research/ai-research-skills

Speculative Decoding

Systems Paper Writing

Moe Training

Presenting Conference Talks

Model Pruning

Outlines

More in AI & Agents

Ppt Template Creator

Skill Development

Plugin Structure

Command Development

Claude Md Improver

Find Skills