Free Speech to Text Online

Convert your voice to text instantly. No upload, no sign-up, no accounts-just speak and transcribe.

🔒 Uses your browser's built-in speech recognition
Word count: 0
Note: This tool requires a modern browser with speech recognition support (Chrome, Edge, Safari, Opera). Microphone access is required and will only be used during your recording session.

How It Works

  1. Allow microphone access: Grant browser microphone permission when prompted. On Safari the transcription runs on-device; on Chrome and Edge your audio is sent to Google or Microsoft's speech service and the text comes back. Absolutool itself never receives or stores your audio.
  2. Start dictation: Click Start and speak clearly. Your words appear in real time as the Web Speech API recognises them.
  3. Edit the transcript: The recognised text is fully editable, correct any errors directly in the text area.
  4. Copy or download: Copy the transcript to your clipboard or download as a .txt file.

Why Use Speech to Text?

Voice dictation is 3 to 4 times faster than typing for most people and reduces repetitive strain from extended keyboard use. The Web Speech API is available in Chromium-based browsers and Safari, providing high accuracy for dozens of languages; Absolutool itself operates no speech backend (your audio is handled entirely by your browser's built-in speech service. Use it to dictate emails, notes, blog posts, and form entries) or to create rough transcripts of audio you're listening to. For accessibility, voice input is essential for users with motor disabilities or those who find typing difficult.

Features

What browser speech-to-text actually does

Speech recognition (also called Automatic Speech Recognition, ASR) converts spoken audio into written text. Modern ASR systems combine an acoustic model (how sounds map to phonemes), a language model (how words and phrases go together in real language), and a decoder that finds the most likely word sequence given the audio. The 2010s revolution was deep learning: neural networks replaced earlier Hidden Markov Models for both acoustic and language modeling, lifting accuracy from roughly 80% on clean speech to 95%+ on cooperative single-speaker audio. By 2022, OpenAI's Whisper demonstrated that a single multilingual model could match or exceed specialized systems across 99 languages.

This tool uses the browser's Web Speech API, the W3C standard for in-browser ASR introduced in Chrome 25 (2013) and gradually added to Edge, Safari, and most Chromium browsers. The API exposes a SpeechRecognition object that streams microphone audio to whichever speech service the browser implements: Chrome and Edge route audio to Google's and Microsoft's cloud speech services respectively, while Safari on iOS 17+ and macOS Sonoma+ runs recognition on-device. Firefox does not implement the Web Speech API at all. This privacy distinction matters: the tool itself runs in your browser and never sees your audio, but Chrome and Edge do transmit audio to Google/Microsoft servers for processing.

For most users, the trade-off versus typing is dramatic. Average typing speed for office workers is 40 to 60 words per minute; average speech is 130 to 150 words per minute. Voice dictation is 2x to 3x faster for getting initial text down, with the caveat that editing afterward is usually still typing. Voice input also matters for accessibility: users with motor disabilities, repetitive strain, or temporary injuries can produce text by voice when typing is impractical. For language learners, hearing whether the system correctly recognized your speech provides feedback on pronunciation. For meeting capture, real-time transcripts help participants and absent colleagues alike.

How this tool works under the hood

When you click "Start Recording," the page creates a SpeechRecognition object (or webkitSpeechRecognition in older Chrome) and calls start(). The browser requests microphone permission if not previously granted, then begins streaming captured audio to the system speech service. The language tag you selected (e.g., en-US, fr-FR, zh-CN) is passed to the service so it loads the appropriate acoustic and language models.

The browser delivers two types of results to the page: interim results (partial best-guesses, updated 5 to 20 times per second as new audio comes in) and final results (locked-in transcription of a complete utterance, typically issued when the speaker pauses for a moment). The tool's textarea shows interim results in a lighter style and locks in final results as they arrive. The word counter updates from the final results only, so it doesn't flicker as interim guesses change. Continuous mode (a checkbox option) automatically restarts the recognition session if the browser ends it after a long silence, which is common on Chrome but rare on Safari.

Once you stop, the transcript stays in the textarea, fully editable. Copy and Download buttons work on the text in the textarea; both happen locally with no server involvement. The tool itself never transmits your audio or transcript anywhere; the only network activity is whatever the browser does internally to communicate with Google's or Microsoft's speech service (or none, on Safari). Your transcript is never stored: refresh the page and it is gone unless you copied or downloaded it first.

Brief history of speech recognition

Real-world workflows

Common pitfalls and what they mean

Privacy: audio handling differs by browser

Unlike most tools on this site which run entirely client-side, the Web Speech API's privacy properties depend on which browser you use. Chrome and Edge transmit your microphone audio to Google's and Microsoft's cloud speech recognition services. Both companies state they do not store the audio long-term for speech recognition queries (as opposed to user-trained voice profiles), but the audio does leave your device, traverses their networks, and is processed on their servers. Safari on iOS 17+ and macOS Sonoma+ runs speech recognition entirely on-device using Apple's on-device ASR, so your audio never leaves your Mac or iPhone. Older Safari versions and other Apple browsers may differ.

Absolutool itself receives nothing. The page calls the browser's speech API, the browser handles the audio (either on-device or via its vendor's cloud service), and only the resulting transcript text comes back into the page. The tool then displays the text and lets you copy or download it; no server call is made by the page itself. For users handling confidential content, the recommended approach is: (1) use Safari on a recent Apple device for on-device processing, or (2) use a dedicated offline tool like Whisper running locally, or (3) accept that Chrome and Edge route audio through Google/Microsoft and use them only for non-sensitive content.

When another tool is the right pick

Other frequently asked questions

Why does the recognition stop after a minute?

Chrome and Edge have built-in timeouts that end Web Speech recognition sessions after about 30 to 60 seconds, intended to save bandwidth and prevent accidental indefinite recording. Enable Continuous Mode in the tool to automatically restart recognition when this happens. The continuous mode introduces brief pauses between sessions (typically less than a second), which may result in occasional missed words at the seams. Safari handles longer sessions more gracefully without timeouts.

Why is the accuracy lower than I expected?

Three factors: (1) Your accent may differ from the training data; consider trying a closer language variant (e.g., en-IN for Indian English, en-AU for Australian). (2) Background noise, microphone distance, and audio quality matter; quiet room and close microphone produce 95%+ accuracy, while noisy environment and distant microphone drop to 70% or lower. (3) Specialized vocabulary (technical terms, proper nouns, brand names) is harder than general speech; for high-accuracy professional dictation, Dragon's speaker training and custom vocabulary are worth the cost.

Can I dictate punctuation by voice?

Not in this tool. The Web Speech API does not interpret voice commands for punctuation; saying "period" inserts the word "period," not a "." mark. Some dedicated dictation tools (Dragon, Apple Dictation, Windows Voice Access) recognize spoken punctuation commands. For browser-based dictation, the typical workflow is: dictate the words, then add punctuation in an editing pass with the keyboard. Modern long-form models (Whisper) often add punctuation automatically based on speech patterns.

Does this work on iPhone?

Yes, on iOS 14.5 and later via Safari. iOS 17 brought on-device speech recognition through Safari's Web Speech API implementation, so your audio never leaves your iPhone. For sustained dictation on iPhone or iPad, you can also use the system-wide iOS Dictation (tap the microphone icon on the keyboard), which works in any text field across the OS.

Why doesn't Firefox support this?

Mozilla has not implemented the Web Speech API in Firefox, primarily due to privacy concerns with the cloud-routing model used by Chrome and Edge, and the engineering complexity of implementing a privacy-preserving alternative. Firefox users on Mozilla's bug tracker have requested speech support for years; Mozilla's official position is that meaningful local speech recognition requires significant resources and they have not prioritized it. For now, Firefox users seeking voice input should use Chrome, Edge, Safari, or a system-wide solution like the OS-level dictation.

Can I transcribe a pre-recorded audio file?

Not directly. The Web Speech API only accepts live microphone input, not file uploads. To transcribe a recorded file, the workaround is to play the audio file through your computer's speakers (or use audio routing software like Soundflower or BlackHole) while this tool listens via the microphone. This loses some accuracy due to acoustic distortion. For high-quality transcription of recorded audio, use a dedicated tool: Whisper (offline, free), Otter.ai, or a transcription service like Rev. For occasional informal transcription, the playback-through-microphone trick works.

Related Tools

Text to Speech

Convert text to audio using neural voices.

Audio Trimmer

Trim audio files to exact length.

Audio Converter

Convert audio between different formats.