How to Convert a Saved Audio File to Text

Supported files and upload limit

The current queued workflow accepts these extensions, provided the server can decode the recording.

• .flac
• .m4a
• .mp3
• .mp4
• .mpeg
• .mpga
• .ogg
• .wav
• .webm
• .aac
• .opus
• .oga

The upload limit is 1 GiB per file. Transcription depends on your available free audio allowance, successful decoding and service capacity. Check available hours before submitting a long recording. Video containers such as MP4 and WebM must contain an audio track.

AIFF, WMA, PCM, MOV, AVI and MKV extensions are not accepted by the current upload flow. Export a supported copy with a trusted conversion tool. Changing the filename extension does not convert a file.

Convert the recording step by step

1. Choose the clearest original recording you have permission to upload. Save it on the device or in a location the browser file picker can open.
2. Open the audio-to-text tool and select one file. Check the upload notice and file details before confirming.
3. Select the spoken language when you know it, or use automatic detection. The interface offers 45 language choices; quality varies with the recording and language.
4. Keep the tab open until the upload finishes. Uploaded jobs continue on the server, so returning in the same browser can recover the result for seven days.
5. Review the draft against the recording, then copy or download the text. Save a backup before clearing browser storage.

There is no fixed processing-time promise. Recording length, queue demand and service availability affect the wait. Preserve the original file if processing fails.

Review the words before using them

Clear speech with little background noise usually needs less correction. Listen again around names, numbers, technical terms and overlapping speakers. Automatic transcription produces a draft, and accuracy is not guaranteed.

For an interview, check each quotation against the recording. For a lecture, verify unfamiliar terminology. For meeting notes, confirm who committed to each action before assigning it to a colleague.

Keep the original recording rather than repeatedly compressing it. A better file extension alone cannot restore speech that was inaudible in the source.

Choose a usable export

• TXT is suitable for plain text and searchable notes.
• DOCX is useful for editing the transcript in a document.
• SRT and VTT are subtitle formats. Their timings are estimates and should be reviewed before publication.

Speaker labels and important wording may need manual editing. Choose a specialist human-review or team tool if that review or collaboration is part of your requirements.

Where audio is processed and retained

Your audio is securely uploaded for transcription. Processing runs on servers and sends audio fragments through Hugging Face Inference Providers to DeepInfra. An internet connection is required.

Audio is deleted after completion or error. Incomplete uploads and queued jobs expire after 24 hours. Private results and incomplete drafts after failure are recoverable for seven days or can be deleted earlier.

Recovery keys and saved transcripts can remain in the current browser until you remove them or clear site storage. The privacy policy explains providers, processing locations, diagnostic retention and deletion. Check it before uploading confidential material.