Speech to Text
Text to Speech
Sound Detector
Three voice tools in one page
This page bundles three small audio utilities. Speech to Text turns what you say into text you can copy, which helps when drafting an email or notes hands-free, capturing an idea quickly, or checking how clearly you pronounce words in a language you are learning. Text to Speech reads typed or pasted text aloud with the voices your browser and operating system provide, so the same text can sound quite different on a phone, a Mac and a Windows PC. Sound Detector shows a live loudness level and waveform from your microphone, which is handy for checking whether a mic works or comparing how noisy two spots are.
How to use each tab
Speech to Text
- Pick the language you will speak from the Language menu (English US, Chinese Mandarin, Spanish, French, German, Japanese or Korean).
- Click the microphone button and allow microphone access when the browser asks. Speak normally; the last few words are provisional and may change as the recogniser settles on a result.
- Click the button again to stop. Use Copy to put the transcript on your clipboard or Clear to start over.
Text to Speech
- Type or paste up to 5,000 characters into the text box.
- Narrow the Voice list with Voice Language, then choose a voice. Adjust Speed and Pitch (each 0.5 to 2.0).
- Press Play to listen, with Pause/Resume and Stop while it speaks. The speech plays through your speakers only; the page does not create an audio file.
Sound Detector
- Click the microphone button and allow access. The large number, the coloured bar and the waveform update continuously.
- Peak and Average summarise the session so far. Press Stop Detection to release the microphone.
What runs where
These features use different engines, and they differ in what leaves your device:
- Speech recognition uses the browser's Web Speech API. The page never receives your audio, but the browser usually does not transcribe on the device either: Chrome streams the microphone audio to Google's speech servers, and other browsers use their vendor's own recognition service, which may run online or on the device. Firefox does not offer this API, so the tab will say it is unsupported.
- Text to Speech uses the browser's own
speechSynthesisengine and the voices listed in the menu. The text is handed to that engine inside your browser and is never sent to ToolHub or any server run by this site. Which voices appear, their quality and the languages they cover depend on your operating system and browser. Voices marked "(online)" are generated by the browser vendor's cloud service, which receives the text; the other voices are installed on your system and speak offline. The page has no server-side speech function and cannot save the audio as a file. - Sound Detector analyses the microphone signal with the Web Audio API entirely in your browser; no audio is recorded or sent.
If you need a voice that is not in the list, add one in your system settings (for example the speech or accessibility section on Windows, macOS, Android or iOS), then reload this page so the browser picks it up.
How the sound level is calculated
On every animation frame (usually about 60 times a second), the detector takes 2,048 audio samples, computes their root-mean-square (RMS) amplitude and converts it with 20 × log10(RMS) + 90, clamped to 0–120. The offset of 90 simply puts typical readings on a familiar-looking scale. The bar turns yellow at 40 and red at 70. Because browsers apply their own microphone gain and noise suppression and every microphone differs, this is a relative level for comparisons on the same device, not a calibrated reading in decibels of sound pressure. Do not use it for noise-exposure or legal measurements.
Frequently asked questions
Which browsers work best?
Chrome and Edge support all three tabs. Safari supports text to speech and, in recent versions, speech recognition. Firefox supports text to speech and the sound detector but not speech recognition. The list of voices depends on your operating system and browser.
How accurate is the transcription?
It depends on the recognition service, your microphone, background noise and how clearly you speak. Choose the language you are actually speaking, speak at a steady pace and use a headset in noisy rooms. Punctuation is limited, so expect to tidy the text afterwards.
Why does recording stop or pause by itself?
Recognition services end a session after a period of silence or after a time limit. The page restarts listening automatically while the microphone button is active, but a short gap can occur. In Chrome, speech recognition usually needs an internet connection.
Why do the voices sound different on another device?
Read-aloud happens in your browser, using the speech voices that your operating system and browser supply. A Windows PC, a Mac, an Android phone and an iPhone each ship different voices, so the list and the sound change from device to device. The page cannot add voices of its own or export the speech as an audio file.
Is anything saved?
No. The transcript, your text and the sound readings live only in the open page and are lost on reload. The page does not use cookies or local storage for these features.