Re: Removing footsteps, creaking floorboards, etc. from dialogue recording

If you have money, ADR "loop" the dialogue and insert it over the original.

If you lack money or the ability to call the actors back in for an ADR session, but have time, work on sweetening the audio you DO have.

Open up the waveform display and you will see the spikes where the noises are. My step number one is to isolate those peaks and edit them down, using the tools in the audio editing app. Often, for a very brief transient, like a bang, some of the good sound will "wrap-around" the noise and the ear will "fill-in" some of what's missing. Step number two is to comb thru the entire recording to see if there are individual phonemes (components of word sounds) you can pick from one section and copy/paste over another.

Like pick up a sibilant "sss" from one word spoken earlier, like "businessss", and plaster a piece of the "sss" over the damaged part of the word "routines". I used to have to do this sometimes in the linear editing days (when this was much more tedious to try and do) and I felt like Gene Hackman in "The Conversation", when I could pull it off, and make it seem like there was never a problem. Had a deal where the speaker had a speech impediment and the producer was really upset the audio was worthless, I told her to take a long lunch and come back. I had plastered one little sibilant "s" into each badly-pronounced word, editing a frame from one tape to another, over and over, the client/producer was astounded: "you found an alternate take!?!?! Where???" "Here" (points to cranium and smiles wryly).

Pulling the right phonemes, with the right inflections, takes time and is not always intuitive: you really have to study the words to see where you can pull a plosive, a glottal, a sibilant. That's hard core EDITING SKILLZ, man:-)

I am FAR from an audio guru, you will want to ask the guys in the dedicated audio forums, but my standard bag of tricks would include notch filtering, compression, muliband Equalizer and downward expansion filters, to try to get some distance between the noise and the voices. Unfortunately, anything you do to filter the noises that is in the same frequency range as the human voices, will alter the vocals a bit. Maybe too much.

A tip to add to the list of "stuff we should have thought about while we were there, and is too late to help us now", is to not just roll some room tone, but to also have the talent sit and just record extra back-up versions of the dialogue, if there is time left to do this.