Lossy Audio Compression => Speech Codecs => Topic started by: schlauf on 2002-10-16 02:39:53

Title: Questions About Speex Audio Codec
Post by: schlauf on 2002-10-16 02:39:53

i just tried to familiarize myself with the new Speex Codec, and have a number of questions about its usage. Sure, I read the FAQ on speex.org and glanced through the speex-search results on HA, but my question wasn't answered there.

Okay, the questions are as follows:

1. What quality levels do the numbers 1-10 represent? What quality level would (roughly) level 10 correspond to - compared with - ogg vorbis (encoded speech material only, of course!). The reason for this question is, that i want to encode some audio material with speech only, but would like to use a highly optimized lossy codec with transparency for this aim. Is this reasonable, or should Vorbis, LAME or <place_your_favourite_codec_here> be used instead?

2. What do the command-switches
-nframes n
-comp n
represent? I mean, I got the point in the manual, but I quite didn't understand what they mean...Which values would I choose for the best quality results possible?

3. Will in (Release Version) 1.0 Bitstream Format finally be frozen? I know it isn't yet in Beta 1.0 ...

Thanks a lot. Oh by the way, if anybody knows: is there a good, freeware (opensource would do the job perfectly ;-) audio waveform editor, preferably supporting VST plugins out there? The only one I found yet is Audacity, and it has proven not to be stable enough yet ...

Thanks a lot for reading & answering!

Title: Questions About Speex Audio Codec
Post by: jmvalin on 2002-10-16 08:07:33
1) The best is always to see for yourself, but if you use the default (quality 8), you get quite good quality (I need headphones and concentration to tell the difference). Quality 2-4 is what I'd consider the bare minimum in quality and 6 is of course medium quality (you probably won't hear much noise if you're using speakers). If you really want completely transparent quality, you can always try quality 10. Note BTW that Speex works only for 8 kHz and 16 kHz sampling rate right now. It may of may not be what you want.

2) --nframes n means that we pack n frames in each Ogg packet (see Ogg doc for details). This is mostly useful at low bit-rates where the overhead of an Ogg packet becomes large. By packing more frames, the total size is slightly decreased.

  --comp n tells Speex how much CPU it's allowed to use. The higher the value of n, the slower the encoding will be, but the higher the quality. (Most useful settings are 1-4)

3) In 1.0 the bit-stream will be frozen. Right now, we're at 1.0beta1 and in a "soft freeze". That means  that it's likely to be the final 1.0 bit-stream. However, there's still a slight chance that we find something that forces us to change it. That's why we say "soft freeze". Also, if there are changes, they're most likely to be backward compatible so that a file encoded with beta1 can be decoded with 1.0final.
Title: Questions About Speex Audio Codec
Post by: schlauf on 2002-10-16 08:17:29
Thank you very much for your response. I love this forum :-)

Happy coding!
SimplePortal 1.0.0 RC1 © 2008-2021