--- title: "Speech to Text - Private Voice Dictation | CanDoYa" description: "Turn speech to text with private on-device Whisper AI. Record your microphone, edit the transcript, and download TXT with no upload or sign-up." url: https://candoya.com/tools/speech-to-text/ locale: en type: browser-tool --- # Speech to Text Category: Social & Creator ## How can I turn speech into text online? Use this speech to text tool to record your microphone and create an editable transcript with Whisper AI on your device. Your recording is not uploaded for recognition. The browser downloads the model on first use, then processes your speech locally with WebAssembly. This tool is built for spoken notes and short dictation, not file conversion. **You control when recording starts, review the take, and then run transcription** in a background worker. Select the language you spoke so the multilingual model can keep its decoding focused. ## How to use it 1. **Record your voice** - Click Start recording, allow microphone access, speak clearly, and stop the take within 10 minutes. 2. **Review and choose a language** - Play the local recording back, then select the language you spoke. 3. **Transcribe on your device** - Click Transcribe speech. The first run downloads Whisper, then WebAssembly processes the audio locally. 4. **Correct and export** - Edit any names or uncertain words, then copy the transcript or download it as a plain TXT file. ## Who it's for - **Writers and students** capturing a paragraph or study note aloud before editing the text. - **Researchers and interviewers** turning a short spoken memo into searchable notes without sending the recording to a transcription service. - **People with typing fatigue** drafting messages or outlines by voice, then copying the result into another app. - **Multilingual speakers** transcribing a short recording with an explicit language hint. ## Expert note Whisper works on chunks of recorded audio rather than converting each word the moment it is spoken. This gives the model more sentence context for punctuation and word choice, but it also means the transcript appears after you stop the recording. Background noise, overlapping speakers, and uncommon names still need careful review. ## Frequently asked questions ### Is my microphone recording uploaded? No. The browser captures, decodes, and transcribes the recording on your device. It downloads model files from Hugging Face the first time Whisper runs, but it does not send your microphone audio or transcript to CanDoYa or a transcription API. ### Is this speech to text tool free? Yes. There is no account, payment, transcription credit, or watermark. Your device supplies the processing power. The practical costs are the first model download, local memory use, battery use, and the time needed to run Whisper. ### How long can I record? Each take can be up to 10 minutes. The limit keeps memory use manageable because the browser holds decoded audio while Whisper works. For an existing interview, lecture, or podcast file, use the separate Audio to Text tool instead. ### Does speech to text need WebGPU? No. The tool uses WebAssembly processing on the CPU for consistent transcription across supported browsers. Processing can take longer on phones and older computers, but the recording remains on your device. ### Which languages can it recognize? The multilingual Whisper base model can recognize many spoken languages. Choose one of the listed languages before transcribing. Accuracy depends on the language, accent, microphone, background noise, and vocabulary, so review important names and quotes. ### Why does the first transcription download a model? On-device speech recognition needs the Whisper neural network in your browser. The first run downloads roughly 150 MB of model data. The browser normally caches it, so later visits can reuse those files unless you clear site storage. ### Can it transcribe an existing audio file? This page is microphone-first and records a new take. Use Audio to Text for MP3, WAV, M4A, OGG, FLAC, or WebM files. Keeping the workflows separate makes each page clearer and avoids putting file and microphone controls into one crowded workspace. ### How accurate is voice transcription? Whisper base produces a useful draft from clear, close-mic speech, but it is not a verbatim guarantee. Noise, music, crosstalk, technical terms, and proper names can cause errors or invented words. Compare important text with the recording before relying on it. ## Related tools - [Audio Recorder](https://candoya.com/tools/audio-recorder/) - [Audio to Text](https://candoya.com/tools/audio-to-text/) - [Auto Captions](https://candoya.com/tools/auto-captions/) - [Language Detector](https://candoya.com/tools/language-detector/) - [Text Summarizer](https://candoya.com/tools/text-summarizer/) --- *Markdown edition of , published for AI agents and other automated readers. Index of key pages: .*