SuperAPI AI Directory
Partner with us中文AI newsFavorites
All toolsOther AI toolsWhisperWeb

WhisperWeb

WhisperWeb is a free AI speech recognition tool that runs locally in the browser, based on the OpenAI Whisper model. Users can transcribe audio and video to text and export it in TXT, SRT, or VTT format without registering, uploading files, or installing desktop software. It supports formats such as MP3, WAV, M4A, MP4, WEBM, MOV, OGG, OPUS, AAC, and FLAC, with a file limit of 200 MB or 20 minutes. Recognition is completed locally on the device, and recordings and transcription results are not sent to the cloud. It offers three models: tiny, base, and small, with base used by default, can handle multilingual speech, and also supports translation into English subtitles. It is suitable for transcription and subtitle creation in scenarios such as podcasts, meetings, interviews, and lectures. Full transcription has currently been tested in Chrome on the macOS desktop version.

Open website

About this tool

WhisperWeb is a free browser-based AI speech recognition workspace, centered on local speech-to-text functionality based on the open-source OpenAI Whisper model. It integrates file selection, transcription review, and subtitle export into a single webpage, and users do not need to register an account or upload files to the cloud.

Uses WhisperWeb is suitable for converting speech in audio or video into editable text and generating subtitles with timelines. It can be used for scenarios such as interviews, lecture recordings, podcasts, meeting notes, and video dialogues, helping to find quotations, verify terminology, prepare notes, or prepare subtitles for video editing.

Main capabilities

  • Local speech recognition: The Whisper model runs on the user's device, and the selected audio and transcription results are not sent to cloud transcription services.
  • Multi-format support: It can process formats such as MP3, WAV, M4A, MP4, WEBM, MOV, OGG, OPUS, AAC, and FLAC, with a file limit of 200 MB or 20 minutes.
  • Model selection: It offers three models: tiny, base, and small, with base as the default; tiny has a smaller download size, and small is an optional upgrade. The larger the model, the more memory it requires.
  • Multilingual recognition: It can recognize speech in multiple languages and output text in the original language, and can also translate into English for free, but it does not support other target languages or dubbing.
  • Subtitle generation and editing: It can generate SRT or VTT subtitles in the original language from audio and video, support editing text and start and end times, and replay segments.
  • Export formats: It supports exporting TXT plain text, SRT, and VTT subtitles.
  • Subtitle format conversion: It can convert SRT to WebVTT, or convert WebVTT to SRT, preserving text and millisecond times.
  • Plain text extraction: It can extract readable text from SRT or WebVTT subtitles, removing sequence numbers, timestamps, and subtitle markup.

How to use

  1. Open the webpage and drag and drop the audio or video file into the designated area.
  2. Select the speech language or use automatic detection, and select the model (tiny / base / small).
  3. On first use, the public model will be downloaded, and subsequent tasks can reuse the browser cache (when the cache is valid).
  4. Recognition is performed locally on the device. After completion, you can proofread the text, edit the timeline, and replay segments.
  5. Export the desired format: TXT, SRT, or VTT.

Note: File processing requires JavaScript support; webpage and model downloads still require an internet connection.

Applicable scenarios

  • Podcast transcription: Convert podcastsConvert recordings into editable text and SRT/VTT subtitles, convenient for use on program pages.
  • Meeting transcription: Convert saved meeting recordings into searchable text without inviting a meeting bot.
  • Interview transcription: Transcribe interview recordings without uploading audio, and replay segments to verify quotes and names.
  • Lecture transcription: Convert lecture recordings into searchable text or subtitles in the original language for easier studying.
  • Video subtitles: Recognize speech in local videos and generate editable subtitles.
  • Format conversion: Convert existing SRT or WebVTT subtitles to each other, or extract plain text.

Limitations

  • Local recognition depends on device memory and computing power; GPU acceleration is preferred when available, otherwise the CPU path is used, and larger files may take longer.
  • Full transcription has been tested in Chrome on macOS desktop; the responsive layout does not mean mobile recognition has been verified. It is recommended to start with short, clear recordings.
  • Direct video links are not supported; local video files must be saved first.
  • Translation into other target languages and dubbing are not supported; only translation into English is supported.
  • There is no cloud project storage; results must be downloaded before leaving the page.
  • Noise, music, overlapping voices, and uncommon names may cause recognition errors; for important content, it is recommended to check against the audio.
  • Accuracy varies by language, accent, and model.
  • Optional usage statistics only record the number of tool uses (start, complete, export), do not record file names, recordings, or subtitle text, and can be turned off at any time.

Related tools

Browse category