Audio editing and hell semantics.

I am working on the edition of audio, by moving, translating, subtitling and testing alternative takes that may match the parts that in the main take were missing or defective.

I have found out that by making video tutorials about video capture, and talking about the audio editing of the resulting video tutorial, the semantics go crazy. This will help me understand and develop with the NLTK Python toolkit. For example:

The audio editing of a video tutorial about video and audio stream recording is almost finished.

That was quite understandable, however there was a lot of word repetition, each word, meaning different things. Also 'video and audio stream' can be linking 'stream' to 'video and audio'; or just 'audio', leaving 'video' semantically linked with 'recording' or alone.

Another more confusing example:

The video tutorial audio editing regarding video and audio stream recording is almost finished.

Modifier nouns are used which turn the associations further complicated.

Recording risks

Screencast narration on a technical subject matter is the hardest that dubbing can get. Although I have never done voice acting, just by allowing you artistic freedom voice acting allows much more space for variances and experimentation. With technical screencasts, like in my case, One has to follow along the many entrenched paths that accuracy encompasses. As the video advances, one has to anticipate what comes next in order to turn the audio experience into something fluid for the audience. Otherwise it sounds forced or inexperienced. Rehearsal helps, but ultimately, mastery of the subject matter treated and experience with microphone and breathing techniques will trump all other virtues.

There are 3 distinct ways of screencasting:

• Direct screencasting. 
• Video first
• Audio first

Direct screencasting
Is simply screencasting while one demonstrates the software. This has the burden of making it all seem fluent and eloquent while at the same time maintaing technical accuracy. The huge advantage though, is that it takes much less time. Audio and video retouching can still be made at post-production.

This is the technique I currently use for fast and short screencasts.

Video first
Allows the video to be recorded in a more laid back way. The audio, however, must match the video by replaying the video and activating the record function of Audacity or any other recorder in a fast move(alt-tabbing may work). Following along, talking about what is displayed in the video is much easier than narrating along with the mouse and keyboard. This allows for a much more professional sounding screencast, at the expense of time consumption trying to make it fit the video, and since the aim is too make it sound more professional, the rehearsal might take longer.

I currently use this for the middle quality screencasts.

Audio first
This technique becomes easy if one has written the script beforehand, and simply reads it along. Once the audio is ready(which is the easy part), the video must be synchronized with the audio(considering the timing based on the current video's framerates in VirtuaDub and milliseconds played in Audacity can be easily translatable.) It is hard to come up with a rhythmic and well timed screencast using this method. It is best left for more elaborate presentations(live footage, animations, composited images and partly screencasts)

This technique I currently use for elaborate, time consuming presentations.

Different video hosting sites for different purposes.

Vimeo is problematic when uploading videos. Not only does it not offer advertising partnership, but it puts ads around one's video without sharing with the content creator! Add to that, the fact that my Zoom demo video is still in processing(which really isn't, it probably simply is timed to be available when it is programmed to, to persuade the standard account user to buy the Vimeo Plus account.) To solve that, I have other accounts in other less known video hosting sites, like Trilulilu, where I uploaded the Zoom tool demonstration for VirtualDub.

YouTube is the ruler, as it offers advertising partnership, allowing independent screencasters to flourish(still pending for me). It also offers from the lowest resolution to the highest of HD, unfortunately adds black strips at the sides if the aspect ratio is not 16:9 .

 ShowMeDo has halted, and is not allowing videos to be uploaded. The uploader is buggy.

Facebook allows video uploading, but the quality isn't high.

I was surprised that Blogspot(Google's Blogger) allows small resolution uploads of videos, with a Flash player reminiscent of YouTube.

The other option is to buy web hosting services, but that is the last option, because one is the payer, rather than the other way around(Like in the case of YouTube ad partnership.)

Zoom tool(motion filter) in VirtualDub

A small illustration of how the zoom tool works. Notice it is a high quality zooming algorithm(Lanczos3).

Watch it on Vimeo for higher quality.

The VirtualDub plug-in is available at Ben Greenwoods homepage.

The VDF file must be placed at the plugins32(VirtualDub 1.10.3) directory















or plugins(in VirtualDub 1.9.11) directory.

Motion Filter tool by Ben Greenwood in VirtualDub

So yesterday I expressed how useless the Zoom 1.2 filter(plug-in) by Donald Graft And Avery Lee for VirtualDub is. 999x999 wasn't a modern resolution. However, there is another tool called the 'motion filter', which allows higher resolutions, and as a bonus, it also allows rotation and panning.

Of course, this is excellent for non-elaborate screencasts in which one needs a fast zoom to a part where one forgot to do it in the screen-capturing session with the excellent tool ZoomIt.

It is always a great idea to rely on various tools as back-up mechanisms for features. In this case a zoom missed in production(with ZoomIt) can be corrected in post-production with VirtualDub and this excellent motion filter tool by Ben Greenwood.

VirtualDub Zoom

Unfortunately VirtualDub's Zoom tool Does not possess the features I am looking for in high quality video zoom tools. The problem, and very serious one, is that the output can only yield a resolution not higher than 999x999! Yes, you read it right, this means no HD!

So back when I created screencasts at ShowMeDo at 800x600 this gave me no problem at all, but now that YouTube accepts the HD options(for people with high Internet bandwidth), capturing the desktop at higher resolutions and rates has become feasable and recommendable(since YouTube's internal processing will offer lower resolutions than the original, for lower Internet bandwidth users).

Which leaves me with GIMP/GAP for zooming.

Crude Video editing

One of my previous posts dealt with the task of creating a video file out of a still frame. That is as low as video editing can get. Perhaps Assembly rasterisation could be the next lowest level.

All this to prove that simple tools(like VirtualDub, MSpaint, and batch or AVIsynty scripts) can become extremely useful when dealing with problems such as a capture program such as Camstudio not allowing screencapturing its own video options menu.

There are other FOSS(free and open source) screencapturing programs, but that will be a left for a future blog post.

For now, learning that VirtualDub will only append equally formatted video files together(this includes the audio bit depth and sampling frequency) is more than enough.

By the way, recording at a faster sampling rate(AKA frequency) and changing the audio to a lower sampling rate, will turn the sound graver, whilst the contrary, that is from a slower sampling rate (i.e. 44.1Khz) to a faster sampling rate (i.e. 96Khz) will create an acuter sound. Thankfully I dealt with silence on those parts, as the narration is appended later in the post-processing of the video.

Planned videos.

I have been working on the screencasting with Camstudio video. Despite the tiredness from little sleep today, at least I will update with the order of the release of the screencasts and video presentations I have at store:

• Internet stream recording with Camstudio
• 'Programmazione in C'
• GIF to AVI with GIMP GAP
• Edutainment promotion video.