Updated: October 25, 2026
| Name | Whisper |
| Overview | Whisper is an advanced AI speech recognition tool designed to deliver high-quality performance through large-scale weak supervision. It supports multilingual speech recognition, translation of supported speech into English, and identification of languages in audio data. Built on a sophisticated sequence-to-sequence architecture, Whisper enhances the process of token representation and prediction decoding. The official repository lists six model sizes—tiny, base, small, medium, large and turbo—with different speed, memory and accuracy trade-offs. Code and model weights are released under the MIT license. |
| Key features & benefits |
|
| Use cases and applications |
|
| Who uses? | Developers, translators, language enthusiasts, and content creators. |
| Pricing | The local code and model weights remain free under the MIT license. Running them still requires suitable computing resources and installation; hardware or hosting may cost money. Separately hosted transcription APIs have their own pricing and terms and are not included merely because the local model is open source. |
| Tags | speech recognition, multilingual support, AI translation, language identification, open-source |
| App available? | The official release is a developer package with command-line and Python interfaces, not a standalone consumer mobile app. Third-party apps may use Whisper under their own pricing, privacy and support arrangements. |
Practical use: Choose a model that fits your hardware and language, transcribe an audio file you are authorized to process, and review the result against the recording. Speech recognition produces text from audio; it is not text-to-speech generation. Treat machine transcripts as drafts, especially where errors could affect people.
Sources: Official models, usage and MIT license; Model capabilities and limitations.
Image context: The existing image shows the OpenAI Whisper GitHub README, its approach diagram and an MIT license label. Its capture date is unknown; it identifies the open-source project and is not a transcription accuracy test or a consumer app screenshot.
