- Home
- AI Video Caption Generators
- Whisper
Whisper
What it is
Whisper is free, open-source software for multilingual speech recognition, transcription, speech translation, language identification, and voice activity detection. It can translate speech from multiple languages into English and offers command-line and Python interfaces for audio transcription. Six model sizes—tiny, base, small, medium, large, and turbo—provide different speed and memory tradeoffs, requiring approximately 1 GB to 10 GB of VRAM depending on size. Whisper runs on Linux, macOS, Windows, and self-hosted deployments, with timestamped output available in TXT, VTT, SRT, TSV, JSON, and JSONL formats. Installation uses pip and requires the ffmpeg command-line tool; Rust may be needed when a prebuilt tiktoken wheel is unavailable. Code and model weights use the MIT License. OpenAI says training used 680,000 hours of multilingual, multitask supervised web data. Performance varies by language, and the turbo model is not trained for translation.
For and against
For
- MIT-licensed code and model weights
- Multilingual transcription and English translation
- Six model sizes with tradeoffs
- Runs on major desktop platforms and self-hosted
- Multiple timestamped export formats
Against
- Requires ffmpeg for installation
- Larger models require more VRAM
- Performance varies widely by language
- Turbo is not trained for translation
The verdict
Whisper offers an open foundation for multilingual transcription and speech processing with local deployment options. Model choice affects speed and memory needs, and translation users must avoid the turbo model.
Who it suits
It suits developers and researchers building speech applications or running transcription locally. Users translating non-English speech should choose a multilingual model rather than turbo.
Plans and prices
Self-hosted Whisper
Python 3.8–3.11 expected · ffmpeg required · approximately 1–10 GB VRAM depending on model
Read 21 Oct 2026 · github.com
Whisper API
25 MiB maximum upload for legacy whisper-1 · streaming unsupported · Free tier 3 RPM and 200 RPD
Read 21 Oct 2026 · developers.openai.com
Whisper (open-source code and model weights)
MIT-licensed code and model weights · install locally with Python and ffmpeg
Read 7 Oct 2026 · github.com
On the record
- API endpoints
- The model documentation lists transcription and translation capabilities and OpenAI API endpoint families including audio transcription.developers.openai.com · read 21 Oct 2026
- API input handling
- The Audio API requires an uploaded file in a supported audio format and does not accept links to audio files.help.openai.com · read 21 Oct 2026
- API model
- OpenAI exposes whisper-1 as a general-purpose transcription model with audio input and text output.developers.openai.com · read 21 Oct 2026
- 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.developers.openai.com · read 21 Oct 2026
- API streaming limitation
- OpenAI states that streaming is not supported with the whisper-1 model.help.openai.com · read 21 Oct 2026
- API upload limit
- For legacy whisper-1 transcription uploads, the maximum request size is 25 MiB.help.openai.com · read 21 Oct 2026
- Automatic language identification
- Language identification is a built-in task alongside transcription and translation.github.com · read 21 Oct 2026
- Company business
- Crunchbase describes OpenAI as developing artificial-intelligence models and deploying machine-learning software tools.crunchbase.com · read 21 Oct 2026
- Company founding
- OpenAI was founded in December 2015.en.wikipedia.org · read 21 Oct 2026
- Company headquarters and size
- LinkedIn lists OpenAI headquarters in San Francisco, California and company size of 1,001–5,000 employees.ai.linkedin.com · read 21 Oct 2026
- Company legal name
- Crunchbase lists the legal name as OpenAI OpCo, LLC and also lists OpenAI LP and OpenAI Group PBC as alternate names.crunchbase.com · read 21 Oct 2026
- Developer extensibility
- The repository points users to third-party web demos, integrations, platform ports and other extensions through GitHub Discussions.github.com · read 21 Oct 2026
- Export formats
- txt,vtt,srt,tsv,json,jsonlgithub.com · read 20 Sep 2026
- Hallucination criticism
- TechCrunch reported October 26, 2024 concerns from researchers that Whisper can insert invented words or passages, including in medical contexts.techcrunch.com · read 21 Oct 2026
- Hardware
- The listed models require approximately 1 GB to 10 GB of VRAM, depending on model size.github.com · read 7 Oct 2026
- Hardware requirements
- Approximate required VRAM ranges from about 1 GB for tiny and base to about 10 GB for large.github.com · read 21 Oct 2026
- 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.freetts.org · read 21 Oct 2026
- 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.frontiersin.org · read 21 Oct 2026
