| |
The Shadow
Registered: Oct 2007
Posts: 304 |
Mp3 to sid
Is there a way to convert mp3s to sids? |
|
... 39 posts hidden. Click here to view all posts....
|
| |
encore
Registered: Aug 2010
Posts: 67 |
algorithm: That last example sounded quite interesting and slightly understandable when she went up one octave in the melody. Does it sound better with double the update-rate or does the envelope-restart (when rising) create too much noise?
Do you use different pulse widths as well? waveform $51 etc? |
| |
Mr. SID
Registered: Jan 2003
Posts: 425 |
Wouldn't this work better if you throw a genetic algorithm at it? Of course it would also take a lot longer... :) |
| |
algorithm
Registered: May 2002
Posts: 707 |
Hi Encore.
The results do sound better at 2x. But there are more issues with noise when increasing rate. (Although i can update the encoder to minimized this)
The example only uses 3 waveforms (triangle, Pulse, noise) and the pulse is of fixed length)
To decrease file size, one of 256 values are used (which in turn point to a lookup table of fixed frequencies and waveforms). The volume table is seperate from this however
To implement the mixed waveforms and full frequencies at some point
The c64 example posted previously is using averaged volume for 3 channels and sounds way off! Will try some methods to reduce the clicking when increasing sustain
The preview output wav is more closer to how it would be |
| |
algorithm
Registered: May 2002
Posts: 707 |
Quote: Wouldn't this work better if you throw a genetic algorithm at it? Of course it would also take a lot longer... :)
Yes:-) this is eventually the idea. Its not using genetic algorithm at the moment, however when implementing the other options, filter, ring mod, i will more than likely use it |
| |
algorithm
Registered: May 2002
Posts: 707 |
The result of using genetic algorithms and shot-gun hill climbing with blind reconstruction..
https://www.youtube.com/watch?v=kD49iTHMm48 |
| |
MagerValp
Registered: Dec 2001
Posts: 1082 |
Here are a few request, it'd be interesting to see how it performs with 2x or 4x:
Kraftwerk: Boing Boom Tschak
Housemartins: Caravan of Love
Nitzer Ebb: Warsaw Ghetto
Anyway, I love this, keep at it! |
| |
encore
Registered: Aug 2010
Posts: 67 |
Really impressive updates, the latest 'Not Gonna Get Us' was waaay more recognizable in the current state.
I'll throw in my requests from irc into here as well. :)
Daft Punk - Around The World
Kraftwerk - Numbers |
| |
algorithm
Registered: May 2002
Posts: 707 |
Thanks for the example mp3's trying it on boom tschak at the moment |
| |
algorithm
Registered: May 2002
Posts: 707 |
Ok, Here is a test example of Boom Tschak
There are still some issues with the encoder that I will rectify in particular the issue with the low frequencies (that has partially been fixed, but still some issues)
The issue is due to the oscillators being free running (the encoder at the moment does not take into consideration the last exact state of the oscillators. This would for example result in a low frequency part being matched well on the encoder, but when played back, it would not line up exactly with the other channels to recreate the digi.
For high frequencies, the error is less as there will be many occasions when the data will overlap.
This encode is an unoptimal encode as I let it run for only 8000 iterations per chunk with 4-shot hill climb. (Much More optimal results would be possible if I increased it to 25000 iterations and 64shot, but would take around 50 times as long
Link to c64 prg here. New sid only. (Will fix this for old sid with frequency interpolation later)
https://www.dropbox.com/s/xyr9s3yd1z6hvtz/testk.prg |
| |
algorithm
Registered: May 2002
Posts: 707 |
A quick few seconds of the daft punk encode.
This demonstrates the issue with the encoding for low bass frequencies.
https://www.dropbox.com/s/g7d5dteyuuh4g04/testd.prg |
| Previous - 1 | 2 | 3 | 4 | 5 - Next |
