I'm sorry if this message is slightly off topic, but I'd like to know an opinion from the speech compression experts reading this list. I'm using spheex to encode human voices at about 15kbps, with very good results. I was wondering if this is the best situation for my setup where I have all the voice pre compressed (not in real time) by very powerful servers with (near) unlimited processing time and computational resources and a client relatively slow (ARM4 processor at about 70MHz). The encoding phase time lenght is not important, I can let it run even for days if it helps compressing the output. The client decodes the data from a static file. My hope was to obtain something of the same quality I'm getting with at 15kbps with a lower bitrate, using more computational time and power on the encoding phase. The audio source are very clean and very high quality. Any hint? Thank you in advance, Andrea P.S. Jean-Marc, I can never say thank you enough for this wonderful software -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20071005/d442442d/attachment.htm
I guess you can always specify --quality 10, but you won't get 15 kbps quality at 11 kbps. Depending on the application though, VBR could be an option. Jean-Marc Andrea wrote:> I'm sorry if this message is slightly off topic, but I'd like to know an > opinion from the speech compression experts reading this list. > > I'm using spheex to encode human voices at about 15kbps, with very good > results. I was wondering if this is the best situation for my setup where I > have all the voice pre compressed (not in real time) by very powerful > servers with (near) unlimited processing time and computational resources > and a client relatively slow (ARM4 processor at about 70MHz). The encoding > phase time lenght is not important, I can let it run even for days if it > helps compressing the output. The client decodes the data from a static > file. My hope was to obtain something of the same quality I'm getting with > at 15kbps with a lower bitrate, using more computational time and power on > the encoding phase. The audio source are very clean and very high quality. > > Any hint? > > Thank you in advance, > > Andrea > > > P.S. > Jean-Marc, I can never say thank you enough for this wonderful software > > > > ------------------------------------------------------------------------ > > _______________________________________________ > Speex-dev mailing list > Speex-dev@xiph.org > http://lists.xiph.org/mailman/listinfo/speex-dev
Hello Jean-Marc, Thank you for your quick answer! I agree with your suggestions, but I was thinking more about pre-filtering or pre-processing of the voice file, before it arrives to the speex encoder. I was wondering if there is some suggested optimization strategy to improve (at least a little) the perceived voice quality while lowering a bit the CBR. Thank you again for your attention, Andrea On 10/5/07, Jean-Marc Valin <jean-marc.valin@usherbrooke.ca> wrote:> > I guess you can always specify --quality 10, but you won't get 15 kbps > quality at 11 kbps. Depending on the application though, VBR could be an > option. > > Jean-Marc > > Andrea wrote: > > I'm sorry if this message is slightly off topic, but I'd like to know an > > opinion from the speech compression experts reading this list. > > > > I'm using spheex to encode human voices at about 15kbps, with very good > > results. I was wondering if this is the best situation for my setup > where I > > have all the voice pre compressed (not in real time) by very powerful > > servers with (near) unlimited processing time and computational > resources > > and a client relatively slow (ARM4 processor at about 70MHz). The > encoding > > phase time lenght is not important, I can let it run even for days if it > > helps compressing the output. The client decodes the data from a static > > file. My hope was to obtain something of the same quality I'm getting > with > > at 15kbps with a lower bitrate, using more computational time and power > on > > the encoding phase. The audio source are very clean and very high > quality. > > > > Any hint? > > > > Thank you in advance, > > > > Andrea > > > > > > P.S. > > Jean-Marc, I can never say thank you enough for this wonderful software > > > > > > > > ------------------------------------------------------------------------ > > > > _______________________________________________ > > Speex-dev mailing list > > Speex-dev@xiph.org > > http://lists.xiph.org/mailman/listinfo/speex-dev >-------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.xiph.org/pipermail/speex-dev/attachments/20071009/8d7721b6/attachment.html