Check an audio recording before transcription

Last updated: October 7, 2026

A rejected upload can come from a damaged file, an unsupported export, a size limit or insufficient recording minutes. Keep your original recording, then check the working copy before starting transcription.

1. Play the beginning and end

Open the exported file in a player on your device. Check that both ends are present and the words are audible. If it will not play, return to the recording app and export or download it again. An interrupted download can leave a file with the expected name but incomplete audio.

For quiet or overlapping speech, keep the original and note the uncertain passage. Converting the container cannot restore words that were not captured clearly. Do not publish an uncertain phrase as a verified quote.

2. Export the format; do not rename it

The audio tool accepts MP3, M4A and WAV. Use the recording app's actual export function and choose a supported format when it offers one. Renaming another format to .mp3 changes the name, not the encoded recording. Open the exported copy again before uploading.

For an M4A from a voice recorder, keep the original export. For a WAV master, preserve that master and use a separate working copy if you need a smaller file. See the MP3, M4A and WAV pages for format-specific problems and playable examples.

3. Separate bytes from minutes

The upload size cap is 100 MB. Your plan's per-file time limit and remaining cumulative minutes also apply; check the current values in the uploader. Guest use is 10 cumulative minutes. A registered account has 30 lifetime minutes in total, including linked guest usage. The free audio guide explains those limits.

A smaller compressed export can help with the size cap. It does not shorten the recording or reduce the audio duration that is billed. For a recording that is too long, export the needed section or review the plan limits. Splitting a recording does not refill the free allowance.

Our same-speech file check

On October 7, 2026, we checked three owned synthetic samples locally. They contain the same spoken script at 16 kHz, mono, with about 7.34 seconds of playable speech:

• MP3 recording: 59,660 bytes, MP3 audio. The MP3 page provides its player and editable example.
• M4A recording: 74,178 bytes, AAC audio. The M4A page provides its player and illustrative note edit.
• WAV recording: 234,816 bytes, 16-bit PCM. The WAV page provides its player and illustrative number edit.

The M4A and WAV were derived from the MP3 with FFmpeg. The WAV is larger; decoding compressed audio into PCM does not recover detail lost in the source. These files demonstrate different encoded sizes, not an accuracy, speed or quality winner. Small codec padding differences are not a comparative billing result.

The three existing local decoding checks passed. The MP3 text below is the original October 6 cloud-demo result. M4A/WAV examples are illustrative manual edits, not new supplier transcription results or recordings from an iPhone or studio. No new provider run was made for this guide.

Please send the revised agenda before Friday. The first session starts at nine, and we should leave ten minutes at the end for questions.

Play the recording to check that wording. In a direct quote, changing “before Friday” to “on Friday” changes the instruction. Writing “9 a.m.” adds information the recording does not provide. If you shorten it into a note, label the note rather than calling it verbatim.

4. Upload, review and save your own copy

Open the audio tool, choose one supported working file and start transcription. If the upload is rejected, use its stated error to distinguish format, file size, per-file duration and available minutes. Do not keep resending the same invalid file.

When text is ready, replay names, figures and passages you will quote. Correct the editable text, copy it or download TXT, and keep the saved file with the original recording. Add speaker names, study notes or summaries yourself; this workflow does not supply automatic speaker labels, timestamps or subtitles.

This short synthetic example cannot establish how a noisy interview, a long lecture or a specific phone export will perform. These steps help diagnose an input; they do not guarantee recognition of every word.

Contact

If you need help with an input, use the VoxText contact page. Avoid sharing private recordings or transcripts in a public support request.