Troubleshooting
Why a word is wrong, and how to improve accuracy
Practical ways to get cleaner transcripts and fix the words that slip through.
Updated 2026-06-14
Speech models are very good, but never flawless. Names, acronyms, crosstalk, accents, and background noise are the usual culprits when a word comes out wrong. The good news is you have several ways to improve things.
- Record closer to the speaker and away from noisy machines.
- Use audio enhancement, with noise reduction and normalization, to clean up steady background sound.
- Add specialized terms to Custom Vocabulary so they transcribe correctly.
- Edit any word inline, and the fix stays with the recording and its exports.
- For imported files, start with the clearest source audio you have.
Fix it once, name it once
When a name keeps coming out wrong, adding it to Custom Vocabulary often solves it for every future recording, instead of correcting it each time.
Related guides
How transcription works, and how fast it is
On-device speech models give you real-time text that works offline.
Edit a transcript, tap to play, and add vocabulary
Fix any word inline, jump to that moment in the audio, and teach Echo Chamber your terms.
Import audio and video to transcribe
Bring in existing recordings, from MP3 to MP4, and get a transcript.
Still need a hand?
Email support@obsidianridgelabs.com, a real person will reply.