Please note: All new registrants to the Panopto Community Forum must be approved by a forum moderator or admin. As such, if you navigate to a feature that is members-only, you may receive an error page if your registration has not yet been approved. We apologize for any inconvenience and are approving new members as quickly as possible.
Audio denoising
Elizabeth Fellendorf
Whiz Kid
Audio denoising would not only improve the audio quality in many recordings, but could also indirectly improve Panopto's ASR feature.
5
Comments
Agreed, essential feature. At the moment it's very torturous to do this yourself, and almost anyone - even with an external mic - will have some fan noise when recording from a laptop- and can be annoying in many cases (esp. since Pantopto seems to use a lot of CPU, so causes fan to whirr when you start using it).
Currently I have to select the option to produce local mp4, upload all the original videos with the panopto app, edit the mp4 locally in Premier Pro to denoise by 60%, export the edited mp4, then go to the online editor, upload a new stream (my modified mp4), go to settings, delete the original stream, then rename the new stream. A real pain! (but final audio quality is much better)
I recall hearing this issue discussed and part of the reluctance to do this at scale was because the de-noising process is computationally intensive and therefore it added expense. I get that.
I think a decent compromise would be to add an option for sessions to de-noise and reprocess the audio for ASR.
To prevent everyone from running it on all the videos, this option could be hidden (so it's only visible if a switch or flag is added to the URL) or protected so users would need to enter a password to run it.
This would allow for some A/B comparisons and broader testing.
I would also love to see something like this, thank you for suggesting it.
@Charles Barbour That is a great option! Have it be similar to human captions.
If the limiting factor really is computation power, I wonder if something could be done in the recorder? Like if there was a "quick edit" process that happened locally to address front/end trims, add audio stuff (denoise and levels), it could even go as far as to set the podcast feed layout.
Thanks for the feedback all, excellent suggestions.
@Anthony Lewis the instructor requesting this feature was indeed concerned about fan noise from his laptop. Your workaround seems a tad arduous, but I'll send it along.
Does anyone know if the new "audio normalization" feature that is included in the Editor (and, IMHO, somewhat hidden) accomplishes this feature request?
Wow, is that hidden!! I was just coming here to create a post asking where is the tool and how do we use it? I was looking for a button or at least an option under the plus button, and support article, which I finally found thanks to searching for normalize audio volume in the above post. Didn't say much unfortunately.
Hi all
Something to note here - denoising won't in itself make poor audio more audible (legible) to ASR and the like, so personally I don't think it would achieve the objectives here - from experience I would argue wherever possible users are recommended to use good quality external microphones placed appropriately wherever possible and understand how to set audio levels accordingly so in that sense an 'auto-gain' setup option within the Panopto recorder might have more merit, e.g. click a button, say a few words and the gain is then set accordingly.
To that end I'd argue that it would be better to promote the use of full headset mics and with that disable any auto-denoising/echo cancellation feature if the recording is being produced via Teams or Zoom - we're having a lot of issues with ASR fails with Zoom recordings in particular. MS are working on an AI based denoising/echo cancellation algorithm for Teams currently in beta which we're told is intended to be a lot less 'destructive' than the current one so if you have the opportunity to try it you might gain some benefits from that.
The Panopto recorder by contrast doesn't need a denoising feature since it's only having to deal with a single source so as a recording tool (particularly if you are generating 'flipped' content), I'd always recommend using the Panopto recorder rather than Zoom since the quality is much better to start with.
TLDR - Normalising and denoising won't and can't 'improve' the quality of your audio, it will only make it more 'consistent' and maybe a little less distracting.
Just to add I think it'd be good for Panopto to update the best practice docs to include notes on good quality headset mics and how to get the best recordings from them, particularly when used in conjunction with video conferencing apps, particularly given the shift from room based recordings to personal ones.