Reduct provides the ability to correct a recording transcript, for cases where the supplied transcript is inaccurate. This may be useful where background noise, poor audio quality, or crosstalk has impacted the transcript's accuracy, or in cases where a synonym or alternate spelling of a name is returned.

Note that the recording transcript should match the audio from your media as closely as possible. If you paraphrase something, the system will not be able to align the transcript with the media, and you will see either gray text, or grey squares. You can choose to skip words when in the Reel editor, which will provide you with more control than directly correcting the underlying transcript.

Select the word(s) you would like to correct with your mouse or arrow keys and click 'Correct Transcript' or tap 'x' on your keyboard.

Edit the transcript to match what you hear, then tap Enter or click Save.

The audio / video files received in a criminal defense context (body-worn camera files, jail calls, dashcams, etc.) often have low quality audio and high background noise, which requires many corrections. To expedite corrections in these cases, we recommend the following best practices.

Select a large selection of the transcript - one paragraph at minimum, and 2-3 paragraphs if you are noticing audio / video synchronization issues. Click 'Correct Transcript' or tap 'x' on your keyboard.

Hit Replay to play the video of the section you selected.

Once you listen to a few seconds, can now Pause the playback, and type.

You can also go forward and back by 5 seconds.
Once you are done with corrections, hit Save.

This is typically the quickest way to correct transcripts en-masse, and also the best way to correct transcripts if there are any synchronization issues between the audio/video and the transcript.

If you correct a single word that appears multiple times in the recording, you will see the option to "Replace all". This will replace every instance of the initially selected word with the replacement, across the transcript. This action can take a few moments, particularly if you are working with a long recording.

If you see gray text after correcting the transcript, that is either because the system is in the process of aligning the new transcript with the audio, or it has not been able to align them. You'll need to type the exact words in the media, to the syllable, for proper alignment to be possible. If you have more than a sentence of misalignment which does not resolve itself after a few minutes, contact support@reduct.video for help.

Stammering can be another major cause for misalignment. The best way to realign your recording transcript is to type out the exact word or syllable pronounced by the speaker, even if the initial AI transcript removed the stammer. For instance, replace "I was a technician..." with 'I, I was a te, te, tech, technician...'

If you later want to cut out disfluencies or stammers, you can do so by using the strikethrough feature in the reel editor.

In addition to correcting the transcript text itself, accurate paragraph breaks that represent speaker changes and changes in topic help make a transcript much easier to skim and to read, particularly in combination with accurately renamed speakers. You can adjust both in Reduct using the relevant tools, which can be found next to the "Correct transcript" button.

In addition to AI transcription, Reduct allows you to upgrade a transcript to human transcription, for human review and higher accuracy. If you expect to use human transcription, you can also supply a transcription glossary for our skilled human transcribers, which can be particularly helpful if your transcript includes specific names, proper nouns, or technical jargon (Note that human transcription is available in English only).