INTRO: You can use CoolEdit to make post-production on the sound. However I ended up not doing any of this for various reasons. But this was the features typically involved in the post production on the sound:
Normalize each channel individually
Remove noise for click-sounds with "Noise reduction". Read help-file about how to use noise-reduction!
In CoolEdit you can do processing of the recorded sound. One is to remove noise. I have done that sometimes but I ended up stopping it since the sound got this "in-a-can" sound over it. Further CoolEdit sometimes created errors in the sound I saved - which was of course useless and confusing!
Filter out high and low frequencies in general.
INTRO: This section contains notes made at an earlier state where I had other ideas of what was the best way to go. I have added update-notes here and there to comment these notes describing my early experiences.
Syncing real time audio
Real time audio: This is good for quick videos, but will make longer videos and lower quality since the voice-over is not so intensive and accurate. You can also record realtime audio of mouse clicks and keyboard and then apply speak later after writing script.
UPDATE: In fact the real time audio is the only way to go to get quality I think. The videos are longer, yes, but that turns out to be an advantage since they are not so paced up and people can follow them like they follow "beamer-demonstrations" they are used to!
Record audio and video in realtime.
Sync Audio/Video in premiere and edit the tracks down, remove mistakes, insert overlays
Export audio signal to WAV, make CoolEdit processing of the two channels and save them as mono MP3s.
Export Video to QuickTime / Animation (see above)
Make new premiere project with QT movie, the click track and speak track (if available): You now have "source material" as QT movie and mp3 sound tracks (with no noise)
Writing script afterward
synthetically speak afterward: Shorter videos. More accurate voice-over from manuscript. More time consuming?
UPDATE: Absolutely a "killer" in a bad sense - don't waste time on this. The scripts will most likely not fit the pace of the videos and the end result is awful.
Capture video with mouse click and keyboard, see above.
Play the video and write speak-text at the same time. Probably you will also find places to remove stuff from video track. (which means it's nice to have video and audio track with moust/keyboard linked here).
Make sure to reality check the text length - does it fit the video time?
Record speak track synchronized with video if possible. Otherwise just record each paragraph with 2 seconds of silence between. Important: As little noise as possible - we need to be able to freely cut the voice over as we like in the pauses.
If the speak-track was recorded simultaneously with video it's easy to import - otherwise it will take some more editing time by cutting it up in pieces.
Recoding audio-dubbings
Recording the original speak track is best done while playing the video at the same time.
After translating the texts they should be recorded synchronous with the original/dummy track. This is easily done by playing the video with original track. The speaker doesn't even have to see the video - just begin speaking each sentence as the English version is spoken (is heard on low volume in earphones). A one-second time delay on the speak is fine - that is corrected when the track is applied.
All texts should be constructed so that there are at least 2 seconds space between each paragraph. Tests has shown that a normal speed speak is 12-14 characters a second.
UPDATE: These guidelines are not necessary. Dubbing is very easy if the original comment track is not too intense. No written translation is necessary. The translation should be done simultaneously in a microphone while playing the video in another application, then merging them afterward.
Result:
Edited QT video track with synchronized click/keyboard sound in and with any overlay and waiting time/mistakes removed; HQ and low size
Speak tracks in MP3
Copyright notice overlay still separate.
Set up in premiere: Ready for additional audio-dubbings
Finalize:
Export audio-tracks (voice over + click) from Premiere and mix them with the video track with Virtual Dub for each language. Use "Direct Stream Copy" for video and "Full processing" for audio, selecting some MP3 sound scheme.
ZIP the whole thing - will bring down the video track FURTHER in size.
Keep archive copies of the Premiere file, and all separate audio/video files in HQ QuickTime - animation / MP3
Since the WMV format is proprietary owned by MS, we can not edit the files with our tools, like Virtual Dub or Adobe Premiere. So we need to convert the files into AVIS/WAVs. This is done by using Graphedit.
Download Graphedit here.
Start Graphedit and go to File->Render media file. Open the WMV file. A new graph is rendered. What we see, are the involved direct show filters, to show this movie on screen.
So we just want to decode the audio and video into a file, we have to remove the 'Matrix Mixer', 'DirectSound Device' and 'Video Renderer'.
After that we add new direct show filters by using Graph->Insert filter. We will need the Direct Show Filters Wav Dest and 'AVI Mux', from the Video Compressors: A lossless codec, like 'Camstudio lossless Screen Capture codec' or 'Techsmith Camtasia Screen Capture codec'. Your screen might look like this now:
Connect the appropriate Filters like shown in this image:
Now press Graphedits 'Play' button. After some time you have the decompressed files on your hard drive. The whole process can take some time – depending on your hardware and the length of your wmv source material.
So – yet you can work with these files in Virtual Dub and all the other applications, that can read avis and wavs. When finished you can feed these files into Windows Media Encoder again, to produce a WMV file.
An alternative way to do this job is to use TSUNAMI TMPEG, http://www.tmpgenc.net/.
We Don't need this feature anymore, but this section is intersting, so I paste it here.With VirtualDub you compress the Audio as MP3 while making a direct stream copy of the video which is compressed good enough.The resulting files are ready for distribution.
Settings for VirtualDub can be downloaded here.
The settings for VirtualDub found in this file linked to above corresponds to this:
Video - Direct stream copy: The video track does not get modified.
Audio - Full processing: The audio track gets processed:
By selecting Compression you select MPEG3 compression, 32 kBit/s, 22KHz, Mono:
When these settings are selected (or loaded from the file provided with the above link) you can use the "Job control..." function in VirtualDub to perform the compression on a whole folder of clips:
After selecting the folder with the clips you want to process (and the folder where the processed versions should go) you start the process:
On my Windows XP the videos compressed with MSRLE looks very disorted. They should look like this ( in comparison, this is totally crisp like screendumps should be):
The reason is not the actual video data inside - that is crisp and clear as it should. The problem is something else and I currently don't know what it is.
The thing is that on ONE of my WinXPs the video is crisp and on two others its blurry. The only difference I can spot is that the two machines with blurry playback are upgraded to Service Pack 1, while the first machine is not. Maybe that does the difference. But trying to copy the codec from the original release over the SP1 codec didn't help. So it may be something else.
Michael Perkhofer: The reason of this can be the color depth of the Windows screen. 16-Bit or 32-Bit. Many Video cards only supprt video acceleration at 16-Bit.
Speak loud. That makes the voice strong and disturbing noises are no longer perceptible. About so, as one in a discussion with 5 people would talk.
Do not forget to breathe. Without air - no loud and clear language.
Work without headphones. Since the recording is still worked over again, headphone disturbs only, since the speaker unnatural hears each breath and then becomes under-consciously jerky.
"Narrate", not holding a speech. Imagine, you present your text a group of people. If you cannot imagine that, set up before you a picture of your dearest ones or something similar before you and tell to it.
Smile! Even if nobody sees it, the result sounds often audibly better, if one shows with speaking mimic.
Speak complete sentences. Even if you can cut later, you should repeat the whole paragraph or at least the sentence before the error in the case of a speech error.
Speak somewhat away of the micro. Chose a good room for audio recordings, then it does not sound like a sheet metal bucket.
Put the micro not directly in speech direction, but somewhat laterally. This will decrease the noises of your mouth, we don't want to record.
Eat or drink nothing 5-10 minutes before.
Most humans have a huge respect and even pros make speaking mistakes repeatedly, if they are faced a microphone. Here only industrious practicing helps to lose around the shyness. That is connected with the fact that one wants to express oneself suddenly selected and error free and besides still tries high-language to speak. You should try (at least with the first attempts), to thus talk like you always talk in every day life, with all the slang etc. When hearing, one notices very fast, what is to be improved (e.g. the pronunciation of the individual words). Also one should not use office- or writing language, do not speak swollen or unnatural. Best is a normal conversation language.
The texts should be spoken briskly, but not fast.
The texts should be examined, whether individual words, or even whole sentence parts do not repeat themselves.
Before the final admission the text should be read out several times loud and a sample admission will show whether everything is optimal, or whether improvements are necessary.
In order to avoid sound artifacts of text sheets, these should be inserted into plastic coverings, or one prepares the text sheets in such a way that the narrator does not need to touch them.
Remember - Typo3 is a great piece of software! ;)
