Audio to Text Converter

Audio to text that knows who said what

A raw transcript is only half useful. If you interviewed someone for an article, you need to know which sentences are theirs. If you are cutting a podcast, you need timestamps to find the moment. If the recording is going on YouTube, you need captions that line up. This audio to text converter returns all three from a single upload: the words, the time each line starts, and a speaker label that separates the host from the guest.

Under the hood, the recording is first converted in your browser to speech-quality audio — mono, 16 kHz — which is all a speech recognizer needs and a fraction of the size of the original. A one-hour video can become a few dozen megabytes before anything leaves your computer, and the video track itself is never uploaded. Recordings longer than about eight minutes are split at quiet moments rather than mid-word, transcribed piece by piece, and stitched back together with correct timestamps.

Turning MP3 to text, video to text, and voice memos to notes

Any format your browser can play works as input: MP3, M4A from iPhone voice memos, WAV, AAC, OGG and FLAC for audio; MP4, MOV and WEBM for video. That covers most meeting recorders, phone dictation apps, Zoom and Teams downloads, camera footage and screen recordings. Converting MP3 to text costs the same as any other format — one credit per five minutes of audio — so a 30-minute interview costs six credits and a three-minute voice memo costs one.

Speaker labels and their limits

The recognizer listens for changes in voice and labels each segment S1, S2 and so on. Rename them in your editor once you know who is who. Diarization works best when people take turns; heavy crosstalk can merge two speakers into one label. For files that were split into chunks, labels are kept separate per chunk (S1, S1·2) so a speaker is never silently mislabelled across a split — easy to tidy in a find-and-replace.

Captions from the same upload

Every audio to text job also produces an SRT and a VTT file. Long sentences are broken into readable caption lengths and each line is timed to the speech, so the files can be loaded straight into YouTube Studio, Premiere Pro, DaVinci Resolve, CapCut or a learning platform. If you only need subtitles for a short clip, this is also the quickest way to get a subtitle generator result you can trust, because you can read the transcript before you publish it.

Looking for song lyrics instead of speech? The lyrics transcriber is tuned for singing and exports synced LRC files. To turn a transcript back into a new voiceover in another voice or language, use the AI voice generator or AI dubbing.

Get a cleaner transcript

• Record with the microphone close to the speaker; distance hurts accuracy more than background music.
• Pick the language manually for short clips, strong accents or lots of technical terms.
• Avoid people talking over each other if you need reliable speaker labels.
• For phone calls, record both sides on separate tracks if your app allows it, then transcribe each.
• Trim long silences and music intros before uploading to save credits.
• Proofread names and numbers before publishing — they are the most common errors in any audio to text tool.

Limits, privacy and refunds

Length and size

Up to 30 minutes per file. Very long files may need more memory in the browser; on a phone, files over 20 minutes are best split first.

Languages

Eighteen languages: English, Spanish, French, German, Italian, Portuguese, Japanese, Korean, Chinese, Arabic, Hindi, Russian, Turkish, Indonesian, Polish, Dutch, Thai and Vietnamese — or leave detection on automatic.

Privacy

Only compressed audio is uploaded; video stays on your device. Transcripts are visible only in your account.

Failed jobs

If transcription fails, the credits are returned automatically, and an interrupted job resumes when you come back to the page.