We generally use Logic Pro X and a Rode NT1A/Audient iD22 for voice. We use FCPx to show the sequence so that the voice can synchronise. We been using a booth cobbled together from two foldable acoustic panels and a SE Reflection Filter X to cut down on the room reflections.

The more I've worked at my desk the more aware I've really become of the limitations of the sound quality. I'd recently read an article about sound absorbers.

RPG Europe's Absorbor come out tops, so I ordered seven panels to give us more options, four for the ceiling and three for the walls.

In essence they've made a great difference, and working with sound is even more of a pleasure.

More here: http://wbavproductions.co.uk/resources/absorbor.html

The short story: 31-OCT, the day the Portuguese Budget for 2013 was approved. People gathered in front of the Parliament to say we're being governed by thieves. Not very original I know, but it's a great place to shoot video and take pictures (well at last until it degenerates into mayhem which is rare around this parts but not unheard of until very recently). This is how I worked it.

1) The extremely simple setup:

GH2, screen folded back as a monitor, Transcend 32GB class 10, Samyang 7.5mm fisheye (my favorite video lens), set and forget to hyperfocal, ISO 400 I believe, SS=1/25s. Flowmotion 2 was used as I never know before hand how much video I'm going to capture or how many photos I'll be shooting, so I need something that is very good quality but doesn't make me have to change cards when something very important is happening. If the card is pre-formatted (the extended, not the quick format) it has never failed me to span.

2) The capture:

This is all about getting there up front and trying not to shake the camera. The police until very recently (and only after a 1h stone shower) has been instructed not to charge, so it's an easy job, you just dodge the TV reporters and other amateurs like me. So, here's my first mistake: the aperture is fine, this lens is sharp as an axe, but should have upped the ISO and set a faster SS because in these situations you are always bound to get bumped and it shows in wobbles after software stabilization.

3) Defishing the Samyang:

Load the MTS into AE. Export as image sequence. Use a batch (http://www.personal-view.com/talks/discussion/comment/69869#Comment_69869) to defish the images and import the output again to AE now as image sequence. The defishing was done using the Panini projection. It's loads more natural on people than Rectilinear, it preserves a lot more corner resolution and still manages to keep a few lines straight. This was the first export and not reasonably quick because the synthesis part takes it's toll: ~2-3fps.

4) Stabilizing:

Throw AE's Warp Stabilizer on the footage. Set the "Smoothness" to 15-25% (it will require much less cropping later) set borders to "Stabilize, Synthesize Edges" (again a lot less cropping) but drop the "Synthesis Input Range" to 0,2s otherwise you'll bog AE. In this kind of footage there is usually little point in using a higher value. Skim the footage looking for black borders and try to resize and reframe it so that they don't become evident. Note: defishing after stabilizing should make more sense as you would have more image to work with (defishing crops it) but from a few rough tries, the AE stabilization isn't clearly as effective when working on fisheye footage. This was the second export and a very long one: below 1fps.

5) CC & Denoising:

Used some of ColorGHear's gears, notably GHrain killer to kill a bit of noise and blockiness, but the footage was already very clean from the start and little color correction was used, just a little bit of levels adjusment. This was the third export and I believe it was even longer than the third due to the noise reduction filters: below 0.5fps.

6) The "so long" part: video effects, tracking and subtitling:

Translations were pretty quick, the hard and long part (at least until I found out how to smooth it) was tracking. It was fastidious to track people's features, takes a long time and needs constant review and adjustment. In the end I found out a few tricks (like increasing the tracking area) and smoothing the tracking points. This eased the job a lot and gave much better results. In the end I had to do it sometimes up to 3x until I was satisfied with it. In the process I had to learn a lot about tracking, smoothing, 3D objects, texting, transitions and light sources. This was the forth, the quickest and last export, probably around 10fps.

7) Audio mixing.

This was the part I was more ignorant about, but turned out to be pretty simple. Even though it's pretty rough, the footage is little demanding with so much crowd noise. It was as simple as importing the last video export from AE into Premiere, importing the original sound, importing also "Le Tigre - Dike March 2001" cutting and pasting a few pieces of this track and adjusting the volume here and there to serve the original events while leaving the words of the protesters audible when needed and off I was to the final export.

The resulting video can be found here:

Here are a few pictures of the event.

Anything more you may want to know about any part of the process I'll be glad to detail it.