Vosk vs Whisper
Vosk
alphacephei.com
Whisper
github.comSide by side
| Figure | Vosk | Whisper |
|---|---|---|
| Price from | Free | $0.01/mo |
| Free plan | Yes | Yes |
| Free trial | — | — |
| Open source | — | Yes |
| Windows | Yes | Yes |
| macOS | Yes | Yes |
| Linux | Yes | Yes |
| iOS | Yes | — |
| Android | Yes | — |
| API | Yes | Yes |
| Self-hosted | Yes | Yes |
| Plans published | 1 | 3 |
| Spec score · AI transcription tools | 8.4 (#15) | 9.5 (#1) |
| Spec score · Speech-to-Text Software | 8.0 (#22) | 8.6 (#15) |
| Spec score · Transcription Software | 7.9 (#31) | 8.6 (#16) |
| Spec score · Speech Recognition Software | 8.0 (#22) | 8.6 (#13) |
| Export formats | TXT,SRT | txt,vtt,srt,tsv,json,jsonl |
| Intended users | The FAQ describes Vosk as a practical, plug-and-play speech recognition library for applications, while suggesting Kaldi for research. | OpenAI describes the open-sourced models and inference code as a foundation for building applications and further research on robust speech processing. |
| License | The Vosk API GitHub repository identifies its license as Apache-2.0. | Whisper's code and model weights are released under the MIT License. |
| Model sizes | Small models are typically around 50 MB and require about 300 MB of runtime memory; large server models can require up to 16 GB. | The repository provides tiny, base, small, medium, large and turbo model sizes, with English-only variants for the first four. |
| Timestamp support | Yes | Yes |
Plans
Whisper
- Self-hosted WhisperFree
- Whisper (open-source code and model weights)Free
- Whisper API$0.01 a month
In their own words
Vosk
Vosk is an open-source toolkit for speech recognition that works offline, including on devices such as Raspberry Pi, Android, and iOS. It supports more than 20 languages and dialects and includes a streaming API. Developers can reconfigure vocabulary; most small models support dynamic changes. Small models are typically around 50 MB and need about 300 MB of runtime memory, while large server models can require up to 16 GB. Language bindings include Python, Java, C#, and JavaScript, among others. Vosk Server has implementations for WebSocket, gRPC, MQTT, and WebRTC, and listed integrations include Asterisk, FreeSWITCH, Jigasi, and UniMRCP. The server can be installed with Docker and configured to use a locally mounted model folder. Listed systems include Linux x86_64 and arm64, Raspberry Pi 3/4, macOS x86 and M1, and Windows x86 and 64 for the Python package. ARMv6 and Windows ARM64 are unsupported; iOS builds are available on request. The API repository lists an Apache-2.0 license.
Whisper
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.