A PRACTICAL NOTE · Local models

How to apply this guide

Play the original in another app before changing the speech model. If playback sounds too fast, too slow or heavily distorted, compare with a fresh recording.

Start with

Make a short standard audio export with a tool you trust, keeping the original. Check the new file’s playback and duration before importing.

Check before you finish

After exporting, check the last few seconds of the new file independently of the first sentence. An exporter may trim a region or change which microphone channel is present. Verify that the clip opens from its final filename before deleting any temporary copy.

Evidence to keep: Write down the speech model, language and app edition; change one of them at a time when comparing results. This page’s key limit is: Play the original in another app before changing the speech model. If playback sounds too fast, too slow or heavily distorted, compare with a fresh recording.

Measure or compare

Make a short standard audio export with a tool you trust, keeping the original. Check the new file’s playback and duration before importing.

Decide what to keep

Compare the same final phrase in both. A clean source is a better baseline than trying successive models on a damaged encoding.

What to watch for

After exporting, check the last few seconds of the new file independently of the first sentence. An exporter may trim a region or change which microphone channel is present. Verify that the clip opens from its final filename before deleting any temporary copy.

Keep exploring