Built on advanced on-device AI and hardware acceleration to deliver professional-grade performance with uncompromising privacy.On-device AI — professional-grade, fully private.
Online Speech to Text Transcriber
Speech to text from a file or your microphone
- Runs on your device, not our servers
- Audio + video in, subtitles out
- Free, no monthly minute meter
Drop a recording here
MP3, WAV, WAVE, OGG, OGA, OPUS, M4A, M4B, M4R, AAC, ADTS, FLAC, MKA, WEBA, MP4, MOV, WEBM, MKV, 3GP, 3GPP, 3G2
Naming the language is more accurate than letting it be detected.
The accurate model is a larger download and noticeably better on non-English speech.
One timing per word instead of one per passage — for karaoke captions and precise editing.
Subtitle files are free here. Both of the tools we are measured against charge for them.
Free: one recording at a time, up to 0 B and 0 minutes.
Why use speech to text?
Dictate notes, draft messages hands-free, or capture rough transcripts without installing separate dictation software.
Benefits
- Hands-free: speak instead of typing.
- Speed: many people talk faster than they type.
- Accessibility: alternative input path.
- Lightweight: runs from a normal web page.
- Export: TXT or JSON for reuse.
Transcribe an audio or video file
Drop a recording in and the whole thing is transcribed on your device — no upload, no monthly minute meter. Video files work too: the sound is read straight out of the container, so an interview recorded as MP4 or MOV needs no conversion first.
What you can drop in
MP3, WAV, OGG, OPUS, M4A, AAC, FLAC, MKA and WEBA for audio, plus MP4, MOV, WEBM, MKV and 3GP for video. Anything else is refused at the file picker with a link to the audio converter, so you never wait through a run that was never going to finish.
What comes out
- Plain text — the transcript as paragraphs, ready to paste into a document.
- Timestamped text — every passage prefixed with its position in the recording.
- SRT and VTT subtitles — the two formats video players and platforms actually read.
- JSON — every passage with its start and end time, for anything you want to build on top.
- One archive for a batch — a queue of recordings downloads as a single ZIP, not one save prompt per file.
Where your audio goes
This matters more than it sounds, because the two halves of this page behave differently and most tools do not tell you which one you are using.
- A file you drop in is processed on your device, not our servers. The recognition model is downloaded to your browser once and runs there. The recording is read locally and nothing about it is sent anywhere.
- Live dictation depends on your browser. Dictation uses the recogniser built into the browser itself, and most browsers send that audio to the vendor's servers to turn it into words.
- Recent browsers can keep dictation local too. Where on-device recognition is available for your language, this page asks for it and offers to install the language pack.
- The page tells you which one ran. The mode shown under the microphone is read back from the recogniser after it starts, not guessed from what was requested.
How it works
Your microphone feeds the browser’s speech recognizer; words stream into the transcript area.
Recognition flow
- Audio: captured while listening.
- Interim: tentative text shown in gray italics.
- Final: committed text appended to the main transcript.
- Language: BCP-47 tag on the recognition instance.
- Stop: ends the recognition session.
Typical workflow
- Language: select first.
- Start: begin speaking after the prompt.
- Review: watch interim vs final.
- Stop: when finished.
- Copy: or export.
Facts
Quality varies by device; Chromium-class browsers are the most reliable for this API.
Key points
- Continuous + interim modes are both enabled in code.
- Changing language while listening stops the session first.
- Word count uses simple whitespace splitting.
- JSON export uses English key names for compatibility.
- Safari/Firefox support may differ from Chrome/Edge.
Best practices
Reduce errors before relying on the transcript.
Tips
- Use a quiet room and decent microphone.
- Match the dropdown to how you actually speak.
- Proofread names, numbers, and technical terms.
- Restart if recognition stops responding.
Ideal use cases
- Notes: quick capture.
- Drafts: first pass dictation.
- Accessibility: when typing is hard.
When not to rely on it
- Certified legal or medical transcription without review.
- Guaranteed offline-only if your browser uses cloud recognition.
Limitations
Browser-dependent API; microphone required; not all languages work equally everywhere.
- Free — one recording at a time, up to 50 MB and ten minutes, with every export format and no watermark.
- Pro — a queue of recordings as one archive, recordings as long as your device can handle, the higher-accuracy model, and word-level timing.
Powered by optimised audio processing algorithms.
Frequently asked questions
How accurate is speech to text for transcribing a recorded interview?
Accuracy depends on audio clarity, microphone quality, accent, and speaking pace. For clear, close-mic recordings in a quiet room, modern browser speech recognition handles conversational English very well. Background noise, strong accents, or fast speech will reduce accuracy, so editing the transcript afterward is always recommended for professional use.
Can I transcribe a pre-recorded audio file instead of speaking live into the microphone?
Speech recognition works with live microphone input. To transcribe a pre-recorded file, play it through your device speakers near the microphone, or use a virtual audio cable to route the audio directly. For pure file-to-text transcription without re-recording, a dedicated transcription file upload tool is more reliable.
What languages are supported for speech to text recognition?
Supported languages depend on your device and its installed voices. Most major browsers support English, Spanish, French, German, Portuguese, Chinese, Japanese, and dozens of other languages. Select your language in the tool settings and speak naturally — the recognition engine handles dialects and regional accents with varying accuracy.
Is my voice data sent to a server when I use this speech to text tool?
The tool uses your browser's built-in Web Speech API. Depending on your device, the audio may be processed locally or sent briefly to the platform vendor's recognition servers (such as Google on Chrome). No audio is stored by this site. For fully offline transcription, use a desktop application with built-in recognition.
Can I use speech to text for dictating long documents or meeting notes?
Yes. The tool supports continuous dictation and displays text in real time as you speak. For long sessions, pause periodically to copy the transcript to your document editor. Auto-punctuation support varies by browser — you may need to add punctuation manually or say punctuation cues like 'period' or 'comma' where needed.
Why does the speech recognition stop or reset after a few seconds of silence?
Most browsers pause speech recognition after a period of silence to conserve resources. Some implementations require continuous mode to stay active. If transcription stops, resume speaking or click the microphone button again. For uninterrupted long-form dictation, keep a steady, moderate speaking pace without long pauses.
from 753 ratings
Rate this tool
Tap a star — it takes a second