Skip to content
SPECIFICTION
Get listed

Whisper

github.com

From $0.01/mo Free plan Open source WindowsmacOSLinuxAPISelf-hosted
Visit github.com
Spec score 9.5 #1 of 99 in AI Video Caption Generators

What it is

Whisper is an open-source speech recognition model for transcription, speech translation, language identification, and voice activity detection. Its documented command-line and Python interfaces process audio files, and it can translate speech from multiple languages into English. The repository lists six model sizes—tiny, base, small, medium, large, and turbo—with different speed and memory tradeoffs; stated VRAM needs range from about 1 GB to 10 GB, depending on size. Installation is documented through pip and requires the ffmpeg command-line tool. Rust may be needed if a prebuilt tiktoken wheel is unavailable. The code and model weights are released under the MIT License, with local installation using Python and ffmpeg. Whisper is listed for Linux, macOS, Windows, and self-hosted deployment. OpenAI says it was trained on 680,000 hours of multilingual and multitask supervised data collected from the web. Performance varies widely by language, and it does not outperform models specialized for LibriSpeech. Turbo is not trained for translation; multilingual models are recommended for translating non-English speech into English.

Spec sheet

What it does
Whisper is a general-purpose speech-recognition model that supports multilingual transcription, speech translation and language identification.Read 21 Oct 2026 · github.com
Purpose
Whisper is a general-purpose speech recognition model for multilingual transcription, speech translation, and language identification.Read 7 Oct 2026 · github.com
Intended users
OpenAI describes the open-sourced models and inference code as a foundation for building applications and further research on robust speech processing.Read 7 Oct 2026 · openai.com
Platforms
The repository gives ffmpeg installation instructions for Ubuntu or Debian, Arch Linux, macOS, and Windows.Read 7 Oct 2026 · github.com
License
Whisper's code and model weights are released under the MIT License.Read 7 Oct 2026 · github.com
API endpoints
The model documentation lists transcription and translation capabilities and OpenAI API endpoint families including audio transcription.Read 21 Oct 2026 · developers.openai.com
API input handling
The Audio API requires an uploaded file in a supported audio format and does not accept links to audio files.Read 21 Oct 2026 · help.openai.com
API model
OpenAI exposes whisper-1 as a general-purpose transcription model with audio input and text output.Read 21 Oct 2026 · developers.openai.com
API rate limits
The documented Free tier allows 3 requests per minute and 200 requests per day, while Build allows 2,500 requests per minute.Read 21 Oct 2026 · developers.openai.com
API streaming limitation
OpenAI states that streaming is not supported with the whisper-1 model.Read 21 Oct 2026 · help.openai.com
API upload limit
For legacy whisper-1 transcription uploads, the maximum request size is 25 MiB.Read 21 Oct 2026 · help.openai.com
Automatic language identification
Language identification is a built-in task alongside transcription and translation.Read 21 Oct 2026 · github.com
Company business
Crunchbase describes OpenAI as developing artificial-intelligence models and deploying machine-learning software tools.Read 21 Oct 2026 · crunchbase.com
Company founding
OpenAI was founded in December 2015.Read 21 Oct 2026 · en.wikipedia.org
Company headquarters and size
LinkedIn lists OpenAI headquarters in San Francisco, California and company size of 1,001–5,000 employees.Read 21 Oct 2026 · ai.linkedin.com
Company legal name
Crunchbase lists the legal name as OpenAI OpCo, LLC and also lists OpenAI LP and OpenAI Group PBC as alternate names.Read 21 Oct 2026 · crunchbase.com
Developer extensibility
The repository points users to third-party web demos, integrations, platform ports and other extensions through GitHub Discussions.Read 21 Oct 2026 · github.com
Export formats
txt,vtt,srt,tsv,json,jsonlRead 20 Sep 2026 · github.com
Hallucination criticism
TechCrunch reported October 26, 2024 concerns from researchers that Whisper can insert invented words or passages, including in medical contexts.Read 21 Oct 2026 · techcrunch.com
Hardware
The listed models require approximately 1 GB to 10 GB of VRAM, depending on model size.Read 7 Oct 2026 · github.com
Hardware requirements
Approximate required VRAM ranges from about 1 GB for tiny and base to about 10 GB for large.Read 21 Oct 2026 · github.com
Independent accuracy test
A September 2026 FreeTTS test measured 11.9% word error for large-v3 with voice-activity filtering and previous-text conditioning disabled, versus 14.8% on defaults.Read 21 Oct 2026 · freetts.org
Independent forensic limitation
A Frontiers study found Whisper best among tested systems on poor-quality forensic-like audio but with about 50% word error, making it unsuitable for forensic use.Read 21 Oct 2026 · frontiersin.org
Installation
The official installation command is pip install -U openai-whisper, with a Git installation option also documented.Read 21 Oct 2026 · github.com
Installation requirements
The repository documents installation through pip and requires the ffmpeg command-line tool; Rust may be needed if a prebuilt tiktoken wheel is unavailable.Read 7 Oct 2026 · github.com
Integrations and examples
The repository links a Google Colab example and invites users to share third-party extensions and integrations in GitHub Discussions.Read 7 Oct 2026 · github.com
Language support
Whisper supports multilingual speech recognition and translation into English, with the repository tokenizer listing available languages.Read 21 Oct 2026 · github.com
Licence
Whisper code and model weights are released under the MIT License.Read 21 Oct 2026 · github.com
Limit: compute and speed
G2 reviewers report that larger models run slowly on CPUs and that long recordings or real-time use can require substantial compute.Read 21 Oct 2026 · g2.com
Limit: setup and interface
G2 reviewers report that setup and fine-tuning can require effort and that Whisper lacks a ready-made user interface and surrounding workflow.Read 21 Oct 2026 · g2.com
Limit: speaker diarization
G2 reviewers report that speaker identification and multi-speaker diarization are weak or absent and often require another solution.Read 21 Oct 2026 · g2.com
Model choices
The repository lists six model sizes—tiny, base, small, medium, large, and turbo—with different speed and memory tradeoffs.Read 7 Oct 2026 · github.com
Model sizes
The repository provides tiny, base, small, medium, large and turbo model sizes, with English-only variants for the first four.Read 21 Oct 2026 · github.com
Model speed options
The turbo model is listed at about eight times the speed of large on the repository benchmark, while tiny is listed at about ten times.Read 21 Oct 2026 · github.com
Performance caveat
OpenAI says performance varies widely by language and notes Whisper does not outperform models specialized for LibriSpeech.Read 7 Oct 2026 · openai.com
Praise: accuracy and languages
G2 reviewers repeatedly praise transcription accuracy, multilingual coverage, accent handling and performance on noisy recordings.Read 21 Oct 2026 · g2.com
Praise: local and open source use
G2 reviewers value running Whisper locally without paid API keys and integrating or customizing it in Python workflows.Read 21 Oct 2026 · g2.com
Processing tasks
The model supports multilingual speech recognition, speech translation, spoken language identification, and voice activity detection.Read 7 Oct 2026 · github.com
Python and PyTorch
The project was developed with Python 3.9.9 and PyTorch 1.10.1 and is expected to work with Python 3.8–3.11 and recent PyTorch versions.Read 21 Oct 2026 · github.com
Rating on G2
4.4 out of 5 from 52 reviews.Read 21 Oct 2026 · g2.com
Runtime dependency
The command-line ffmpeg utility is required and installation examples are provided for Ubuntu, Arch, macOS and Windows.Read 21 Oct 2026 · github.com
Timestamp support
YesRead 20 Sep 2026 · github.com
Training data
The model was trained on a large, diverse audio dataset collected under weak supervision.Read 21 Oct 2026 · github.com
Translation
Whisper can translate speech from multiple languages into English.Read 7 Oct 2026 · openai.com
Translation limit
The turbo model is not trained for translation tasks; the repository recommends multilingual models for translating non-English speech into English.Read 7 Oct 2026 · github.com
Usage
The repository documents command-line and Python interfaces for transcribing audio files.Read 7 Oct 2026 · github.com

Plans and prices

Self-hosted Whisper Free MIT-licensed code and model weights; local installation Python 3.8–3.11 expected · ffmpeg required · approximately 1–10 GB VRAM depending on model Read 21 Oct 2026 · github.com
Whisper (open-source code and model weights) Free MIT-licensed code and model weights · install locally with Python and ffmpeg Read 7 Oct 2026 · github.com
Whisper API $0.01 a month $0.006 per minute of transcription 25 MiB maximum upload for legacy whisper-1 · streaming unsupported · Free tier 3 RPM and 200 RPD Read 21 Oct 2026 · developers.openai.com

Every Whisper plan and price →

Strong points and limits

Strong points

  • MIT-licensed code and model weights
  • Command-line and Python interfaces
  • Six model sizes are listed
  • Supports transcription and language identification
  • Can translate speech into English

Limits

  • Requires the ffmpeg command-line tool
  • Models need approximately 1 GB to 10 GB VRAM
  • Turbo is not trained for translation
  • Performance varies widely by language

Who it suits

Whisper suits developers and researchers building speech-processing applications or conducting further research. Its local interfaces and model-size choices also suit users able to manage the stated software and hardware requirements.

Verdict

Whisper provides a free, locally usable toolkit for several speech tasks, with model choices that trade speed against memory needs. Translation users should select a multilingual model rather than turbo for non-English input.

Where Specifiction places it

Alternatives to WhisperAll alternatives →