Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 1 | These wav files show how Sonic performs at increasing speech rates. All sound |
| 2 | sampels are in the public domain. |
Bill Cox | db3f14d | 2011-04-15 12:43:20 -0400 | [diff] [blame] | 3 | |
Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 4 | sonic.wav |
Bill Cox | db3f14d | 2011-04-15 12:43:20 -0400 | [diff] [blame] | 5 | This is a sonic 2X sped-up version of a public domain librivox.org recording, from |
| 6 | the audiobook "Princess of Mars". |
| 7 | |
Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 8 | soundtouch.wav |
| 9 | This is the same recording as sonic.wav, but sped up using soundtouch, which |
Bill Cox | db3f14d | 2011-04-15 12:43:20 -0400 | [diff] [blame] | 10 | uses WSOLA rather than the sonic algorithm. Even at 2X speed up, you should be |
| 11 | able to hear the characteristic WSOLA distortion relative to the sonic version. |
Bill Cox | 9e7f1ae | 2010-11-04 09:19:09 -0400 | [diff] [blame] | 12 | |
Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 13 | talking.wav |
Bill Cox | 9e7f1ae | 2010-11-04 09:19:09 -0400 | [diff] [blame] | 14 | This is my father talking, using a decent microphone and 44KHz sample rate. |
| 15 | |
Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 16 | talking_2x.wav |
Bill Cox | 9e7f1ae | 2010-11-04 09:19:09 -0400 | [diff] [blame] | 17 | This is his voice sped up by 2X using Sonic. |
| 18 | |
Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 19 | espeak_s450.wav |
Bill Cox | 9e7f1ae | 2010-11-04 09:19:09 -0400 | [diff] [blame] | 20 | Sonic also performs well at increasing the speed of synthesized speech. |
Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 21 | espeak_s450.wav was generated using 'espeak -s450 -f test1.txt -w |
| 22 | espeak_s450.wav'. This is the highest speed currently supported by espeak, |
Bill Cox | 9e7f1ae | 2010-11-04 09:19:09 -0400 | [diff] [blame] | 23 | though Sonic can speed up espeak to much faster rates. |
| 24 | |
Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 25 | espeak_sonic.wav |
Bill Cox | 9e7f1ae | 2010-11-04 09:19:09 -0400 | [diff] [blame] | 26 | This was generated with 'espeak -f test1.txt -w out.wav; |
| 27 | sonic 2.6 out.wav espeak_sonic.wav'. Sonic sped it up 2.6X, which is about the |
| 28 | same speed as espeak at -s450. I personally feel that the sonic sped up sample |
| 29 | sounds better than espeak at -s450. |
| 30 | |
Bill Cox | ec23ae0 | 2011-10-21 03:02:47 -0400 | [diff] [blame] | 31 | twosineperiods.wav |
Bill Cox | db3f14d | 2011-04-15 12:43:20 -0400 | [diff] [blame] | 32 | This is just two sine periods, which is too short to hear. However, it's |
| 33 | useful for making sure the flush function works correctly. A 2-X speedup should |
| 34 | result in one sine period with no distortion. |