What a difference "clean sources" could make!
Playing with new version of x264 (r341) and clean DV25 sources it become quite obvious that x264 could be pushed bellow 500kbps@30fps (but still above 400kbps) and thus beat ASP head on.
Since one cannot go significantly bellow 500kbps, the only option to stream SD at bellow 0.5Mbps is to encode at lower frame rates. In most cases 15 fps is quite adequate. This brings us to 200-250kbps for video that translates into less then 1MB per 30 sec. or 2MB per minute.
At those rates allowing more then 10-15% for audio is a waste, thus audio too needs to be resampled at 22KHz and AAC encoded at around 50%-75% quality (bellow 50% audio is junk). The following table lists bitrates per quality settings for 22KHz when encoded with faac
| Quality |
bitrate |
K/min |
| 100% |
43kbps |
320 |
| 75% |
37kbps |
280 |
| 50% |
28kbps |
230 |
Video rame rates could be converted (deinterlaced, denoised and scaled) using simple Avisynth script. Audio could be resampled by Audicity before beeing fed to faac. Note, then Audicity adds junk to the end of file when converting from 48KHz to 22KHz, so verify before feeding to faac.
All in all, x264 could squeeze upto 10 minutes of SD video into 20MB.