When you show people recordings of themselves, they are inevitably appalled and overly critical of their performance. I am no exception. When it came time to edit my presentation from Northern Voice, I was not happy with the number of um's and ah's. I vowed to edit them all out. The ostensible reason is that it would make the presentation more pleasant to listen to, but it was probably mostly just vanity.
But there was some technical interest too. How to edit um's out of video was asked during my presentation and after some fumbling, we came up with the answer: paste some room tone over the um's. Although I have done tons of this kind of editing for audio, video is different: you can't just cut it out or the video will have a visual discontinuity that would be jarring.
So I took my own advice, looked for a quiet passage in the audio and saved it out to a separate file. Whenever I encountered an um, I would copy a section of the appropriate length from this file and pasted it over the um. By doing a little crossfade at the edges of the paste, the edit was audibly undetectable. And since your mouth doesn't move much when you um, you don't really notice that the sound is missing.
The concept is straightforward but the procedure quickly becomes very tedious. I used Adobe Audition for the edits and wrote a script to automate the process. Audition's scripting capabilities were too limited for the job so I used an absolutely stunningly wonderful scripting program called AutoHotkey. The end result was very nice. You select the region of audio containing the um, press a button and the rest happens automatically. It even leaves the cursor in position to play some preroll so you can hear how the edit sounds in context.
The script proved so easy to use that I also cleaned up breathing sounds and other intrusive mouth noises. I will post the script to our website eventually, but if you're in a hurry drop me a note and I'll send it to you.
Most of the rest of the time on the audio was spent cleaning up the interference on the wireless mics. Ugh. And I brought the audience Q&A sound up so you could hear it better. One test viewer said it was still too hard to hear, so I put brief summaries of what the audience was saying in text in the lower third.