Editor and self-learning

How to review a transcript quickly, and why each one is more accurate than the last.

The editor: doubtful spots stand out

No need to reread the whole transcript. The editor walks you through the places where recognition may have gone wrong.

Editor

Words the recognizer is unsure about are underlined; lines where the voice doesn't clearly match the speaker are boxed.

One button jumps to the next questionable place and plays that snippet right away.

Click a line's timestamp or put the cursor in a word, and you hear exactly that moment, slowed down if you need.

Reassign a line to another speaker, split or merge lines, insert a mm-hmm or a pause, turn a period into a question mark.

A feed of lines, the player right under your thumb, every action a tap away. No mouse needed.

Lines you've listened to or corrected are marked automatically, so you always know how much is left.

Every reviewed recording makes the next one more accurate - for you, your terminology and the people you talk to.

Learning uses only the corrections you've confirmed and only your own voice. Training data is deleted along with your profile.

This is what a TrueTranscribe transcript looks like

Correct a word the same way twice, and it gets fixed automatically in future recordings: names, terms, method names.

Your terms are fed to the speech recognition model in advance, so it gets them wrong less often.

Based on your decisions, the service learns which short voice overlaps to mark as talking at the same time.

With your consent, the service remembers your voice and recognizes you right away in new recordings. Other speakers' voices are not stored.

Upload a recording and see how much of it a regular transcript would have lost.