Episode 0.4 was the first episode recorded with a proper digital camera. It was a totally different experience to shooting with my Olympus C-8080. I think the most important factor was that I wasn't afraid of going over time. The C-8080 was limited to about 25 minutes of video at high resolution, and this was always in the back of my mind. Now that I have 588 minutes of free time on the camera, I'm not so worried anymore ;)
I decided that I would try out the built in camera for the audio recording of episode 0.4. For a start it was a stereo mic, and secondly, it seemed to perform well during my early tests. However, when trying to use it in an actual recording environment, it turned out to be a no go. The signal to noise ratio was way too extreme. Luckily I left nothing to chance and I still used the trusty N810 to record onto as well. I have now also purchased a tie mic, so it will be interesting to see how this changes things.
Recording went well, and it was a totally different experience to go and record on location during the watercooling segment. I was now, not in a familiar environment, and the audio took a fair amount of cleaning up, in order to get it into a usable format. Luckily Audacity and my usual recipe helped out here.
Another first for episode 0.4 was the new 3D overalys. Although they don't look like anything special, it took a long time to find out a way of doing exactly what I wanted. Starting with Blender, I created a cube, and animated it so it spun and moved in and out of the picture. Then I placed a textured plane which would house the white text. The interesting part was getting the cube to wipe the text in and out. The image had to be saved as a PNG, with a transparent background. Now the hard part, how do you texture something to obscure something else, but at the same time be invisible? You turn it's alpha down to 0, but don't turn on transparency, therefore it is never seen, but it also cannot be seen through.
Now because Blender cannot export video with a transparency layer, I had to employ a fairly special workaround to get these overlays into Cinelerra. I started by exporting a sequence of frames out of Blender as PNGs, and switched on the RGBA mode in the renderer. This left me with 250 seperate PNG files. Next, I changed the options in Cinelerra to allow me to import images as single frames, and opened the entire set in one go in a new project. This was then exported as a Quicktime for Linux file, but using the Uncompressed RGBA compression format. When I loaded my real project, I was then able to drag these files in. The only cost of doing this was a) a little time and b) a bit of disk space. Each overlay takes around 300Mb. Making for a total of over 2Gb for episode 0.4
The show ended up being a lot longer than originally hoped, going over the 30 minute mark by a further 15 minutes. I have since heard someone state that 45 minutes is just about the right length for an IPTV show. I guess only time will tell.
I had received a great email from Rolf from the meetthegimp project about using Auto-Duck for segments which had music backing tracks. I tried this out, with the desired effect being the music would lower in volume when I was talking, and come back up again when I stopped. Whilst it was a great idea, I think the subtle difference between meetthegimp and progbox may lie in the way I talk ;) I don't tend to stop much, trying to cram in as much information as I can. To begin with the fadein/out time was set to 0.5 seconds. This proved to be annoying, as the music would jump up and down very quickly. So I tried changing this to around a second, and I found there were almost no points in the audio where I was quiet enough for the music to increase in volume. Coupled with the fact that it added an extra complexity layer to editing, I decided to manually pull up the audio in segments where I was quiet for a reasonable amount of time.
The editing was a little different this time. Not previously having the space to record large segments this time I decided, both from a trying the camcorder for it's audio quality and from an ability to fade from screen to face easily, to record myself whilst I was demoing on screen. Episode 0.3 was dubbed audio well after the screen casts were shot, and I'm still trying to figure out which is best. On the one hand you get a totally immersive experience as I'm talking whilst I'm typing, and also I tend to go slower. On the flip side, the audio is much cleaner when dubbed afterwards and you don't hear the keystrokes. As it turned out this fading from screen to face mid segment, never happened and only really marginally helped in the editing.
Cinelerra held up really really well this episode. I have had many people ask me about it's stability. In truth, the only time it really is unstable, is in importing the videos in the first place. It seems to want to try to read a file before it's complete, but this was really only a minor inconveinience and other than that, I had only one crash during the entire 5 hours of editing.
0.4 also posed a challenge, because all of the screengrabs were recorded in different resolutions. In the end, the final episode file had around 10 audio tracks and 10 video tracks. I again, used the projector function to zoom these tracks, which, overall worked really well.
Overall, I'm happy enough with episode 0.4 that I can officially close off series 0.