Two additional issues with transcription
Hello,
Have come across two additional issues with the new transcription feature.
One — When there is a significant gap in time between clips, the the start of a clip will be transcribed as expected, but the track marker will be put at the end of the preceding clip. Please see screenshot.
In the screen shot, "Because we have all the votes" should actually be associated the with the very NEXT clip. Same for "I don't think we can never..."
Two — In addition to sometimes creating dozens of "BLANK AUDIO" track markers (previously reported a few days ago), the transcription occasionally repeats a word dozens of times. In this case, it is the word "you" repeated every couple of seconds for the entire length of the transcription. Once this happens, the transcription never corrects itself.
With these three issues (including the BLANK AUDIO issue) I'm unable to use this in regular production under deadline. But it is SO close! Looking forward to seeing where this goes.
Thank you!
Rob
- SSoundFlow Bot @soundflowbot
Thanks for posting a question or an issue related to the 'Notes' package.
This package is made by @chrscheuer. We're auto-tagging them here so that they will hopefully be able to help you.Please note, that the best way to get help with a script, macro or other content installed from the Store is to select the script you installed, then click the red Need help button, and then click "Get help with this script or macro".
By using this workflow, the developer of the package will get access to more information so they'll be able to help you quicker.
You can read more about how to best get help in this article: bit.ly/sfscripthelp - SIn reply toRob_Byers⬆:SoundFlow Bot @soundflowbot
This issue is now tracked internally by SoundFlow as SF-1464
- AIn reply toRob_Byers⬆:Andrew @Andrew4
Hey Rob,
This may or may not be helpful but I'll share in case it is. I was having some issues with the transcription/track marker feature as well. Not exactly the same, but similar. I found that playing with different language models gave me pretty significantly different results. For example I was getting tons of bugs (to the point the transcriptions were unusable) with the large model but have now been using the Small (English) model for the last month or two with really good results. If you haven't messed around with that, I would definitely recommend at least giving it a shot.