pouët.net

After the Just Dance demonstration, some people had asked whether it was a possibility to stream and decompress at twice the sample rate. This demonstration shows the possibility of this. Sample rate is at approx 7.812 and there are no interpolation tricks (eg half sample rate data) I have used another one of my audio encoders to compress the data for this demonstration.

The compression method for the audio is based on Vector quantisation where segments of extremely small chunks of audio are analysed and the 'best few' which would represent the entire audio are kept.

256 'optimum' waveforms are selected and then modified using clustering. I have quoted the word optimum as the encoder can still be improved immensely.

Krills loader just about manages to keep up with the decompression. Comments and critisisms welcome. I would like to see whether the streaming works well for most drives.

There is a known bug where the audio sounds distorted when running the demo sometimes (should have fixed it) But 3am in the morning and work in a few hours...

Wow.. great! you even make it loop.. so no need to re-start manually. :) Hum.. yep, I got the demo to pause a little bit which create a bit of sounds distorted and comeback. I am trying with Vice 2.2 and sid with resid-fp. Sorry, I don't have real hardware to try out.. but if I will, it will be on NTSC anyway. ;)

Btw, we were talking about you last saturday at a mini-demo held in Montreal.. when PsychNerd from SceneSat Radio interview us. Here is the link :)

Running back Just Dance 64 on the emulator.. I do prefer that one. I think it's because of the song. btw, TronicSynth where telling me that you probably have a sid channel or 2.. running a remake of the song in .sid. I don't think so...

btw, here is a mix of 'don't worry' Sanyo (aka TronicSynth) did for us back in 1991 for our disk demo named Evolution by Optic Vision (it's a ntsc demo). It's a 4-bit samples and I believe he did sample around 7-8kHz. Our NMI playing routine was handling start-end sequence allowing him to edit a sequence order of samples to play. I think it would be awesome to have this capability in your streaming audio vq decoding... (if possible?). :) Congratulations again for these awesome release!

Algotech is crazy

Awesome :D

The demo does not run well with Vice. (it freezes and distorts the audio whenever the next audio section has started loading. I would recommend either Hoxs64 or a real C64. As with all demo's which use the 8 bit technique, The minimum of Resid-FP is required for Vice. Hoxs64 works without any issue. Had a look at the scenesat radio link but it is over two hours long. Can you give me the area when the audio starts.

I have created a in memory sequencer which can sequence audio sections (and decompress these on the fly) Doing this while loading is on the agenda as well :-)

Is this background "noise" from the emulator? Otherwise the audio qualits is nice! :)

Oh, I will try Hoxs64 when I have time.. I do not know this emulator. For scenesat, hum.. well we switch in French (sorry...) from 01:42:30 to 01:45:20.

I guess for memory sequencer, you will need to have this sequence in the very few kb ram of the drive.. so the drive can handle where to load? Or perhaps you can have that sequence in a specific sector-track of the drive.

i want yootoob

Sick stuff :D

Going home to the US in two weeks; I will gladly test this on NTSC if you'd like.

The memory sequencer is very straightforward and just pulls out the required data and feeds it through. Disk loading based sequencer would require more planning with the audio in regards to buffers. but certainly doable. All this would be done without any drive code or info written to tracks on the disk. I may at some point release my pc based encoder (which can also be used for other machines)

rulez

Unfortunately the audio noise (that sometimes occurs) when the demo is run is a bug of the program itself and not the emulator. NMI occurs every 2 raster lines, but it does not always start from the beginning of the rasterline which causes distortion. Will fix it in future production (which will hopefully feature 16:1 compression at the same quality :-) The logo should have also had each even frame shifted one pixel right for hires effect. aaargh.

Humm.. I wonder if a tool like that could be use for GP2X. The hardware can decode a mp3, but that take lot's of cpu resources, currently we go with a ported mikmod...

Great sound

Not much to see but much to admire. Nice achievement. I am looking forward to some practical use of it.

The gp2x would not have any issues in decoding considering that it can be done on a 1mhz c64. However for higher quality I would recommend other decoding methods which would be doable on a gp2x but not a c64.

In my case I have some issues on real hardware and no problems on Vice 2.2. Or maybe it runs incorrect on both? Well, I've got to check my hardware and transfer methods.

Totally awesome!