Blackmagic Forum

This didn't really improve within the last updates. Issues:

- Parts of the spoken words are just randomly left out. Even if voice and audio quality doesn't change at all, a few sentences will just not show up in the transcription. The transcription is pretty much useless, if it leaves out parts.

- If there are multiple speakers, like interviewer and interviewee, the AI doens't recognize that. Seems straight forward to me to distinguish between two different voices, but it's not happening. You have to manually structure the text.

- Those "(...)" are usually totally random, even within a flow of speaking. Just de-structures the text result.

This all leads to either manually editing the text for quite some time until you can finally send them to a client or work with them. Or you simply export the audio and use a better transcripts.

I like working with DaVinci Resolve. But this feature is frustrating. We know tech is already ahead of what we get here. And we seem to be close to something that speeds up our workflow. But right now, it's not there to really save any time.

Thinking outside the box....is there something in the menu to slow down the speed it transcribes. Maybe there is something I can do to make the transcription clearer. It's not doing a good job writing out the correct words.

ai transcription: "(...) I would want to get a living over America, and..."

from the correct sentence: "someday, I would like to step away from corporate America"

That's way off and not worth the time to manually edit.

Yeah this is still an issue. Hard to be confident clicking the remove silent parts button when it often eats big chunks of dialogue. Hoping for some improvements here!