pouët.net

Finally got back home at 3am this morning. Had such a wicked time, everyone was really friendly. The Real-Time coding compo was really impressive and in fact probably the highlight of the whole party. All the other compos were great as well, although I did find myself totally baffled by the Wild comp.

The Friday night live acts were awesome too. Karen Novtny was cool, glxblt played a blinding set on what can only be described as a f*&ked DJ mixer! and the metal band made me feel like I was 13 again!

Apologies for the slackness of Radio broadcast on Saturday, was getting a bit too drunk and tired. Will try and be a little more professional next time ;)

Special thanks to Azzaro of MAWI for letting me sleep in his room at the Hostel and also apologies for snoring! Also would like to thank Tom Dagone for coming on the trip with me and making sure we got to where we needed to go. I could speand hours talking to theres guys.

Also big ups to the organisers, didn't think it was any more disorganised than most parties I've been too and they seem to get everything running without too many delays.

I would recommend everyone goes to WeCan. The people or so friendly and know how to have a good time. Great party all round.

That was outside perspective. ;)

What really happened was: "F**k, some effect attached to the pads don't work. Are all pads really working?" And then I jumped to test myself as a last resort.



You asked for it. ;)

The idea was utterly cool, and Bonzaj highlighted all the awesomness, so I'll focus on the darker side.

The FFT input was complete crap. You got 128 FFT values from what looked like 128-sample FFT. With NO filtering. And there was no history or whatever so single shader had no means of filtering it (no to mention time taken). To make things worse, the Titans were running with no VSync, so any use of FFT resulted in massive tearing.

Pads were better, but not by much, as they also didn't have any filtering. Just 0 or 1 depending on whether pad was held or not.

What I would really like to see, is filtering/peak detection on FFT (probably a lot of tweaking required to make it usable) and release on pads, e.g. instant 1 on note on and 1 second decay to 0 after note off.

And last but not least, the voting system was weak. Usually we had clear cuts which it of course handles, but Bonzaj mentioned two cases where it didn't work. Using partymeister for final round was brilliant, but Reg vs w23 was quite close call because Reg was having much more stuff going on at that exact moment (switching of literally several dozen of various 2D filters applied to a photo of DeadMan) than w23, who at the end of first round had diffuse-lit sphere. I have no idea what Bonzaj would do if w23 and me met earlier than finals. :) Looks like proper tournament tree preparation is vital for this compo.


When it comes to further ideas on how to extend the compo environment:
1. If we could write the shaders in DirectX 11 I would LOVE to be able to define and use several UAVs. This would enable some serious effects with frame-to-frame persistence.
2. Thinking about VJing, it would be nice to have a nano-kontrol (in addition to remote pads) for each participant - for testing, tweaking and some VJing during final presentation (which could be staged as a final deciding battle - VJ to the music and orgs videomix both shaders).


And while we're talking about the compo, much much respect and congratz to w23.



Seriously, don't think about having a single slot. You need separate one for each stage with breaks in between, so coders have at least half hour to regenerate and cool their brains. The tension during the compo can only be compared to the one you feel fixing a crashing demo 20 minutes before compo deadline. And you get this three times. ;)

Fortunately enough, a large beer after the last round is a really good brain coolant.

All the stuff from WeCan, as well as results are now in the upload area at scene.org and should become public as soon as the admins notice.

All game entries have not been uploaded as they were disqualified for having been released at previous parties.

Only photo-based wild entries have been uploaded; all the remaining productions used commercial sound-tracks that scene.org admins would presumably not enjoy particularly much.

Elude and Futuris' entries have not been made available as per authors' request.

@kbi: My entry (TimeLoop) and SuperHOT, too, as far as I know were previously released, but not at parties. But I still agree that the official rules* were a bit lax and led to a general outburst of microwave reheated entries. ;)


*) "Entry is allowed to have been shown earlier but not on a demo-scene party."

KK: Aceman is responsible for entry acceptance, so please contact him in case you have any doubts :)

kbi: As far as Bonzaj told me, org team had 1/2h discussion about this, so this was late change of rules (but a wise one) as far as I understand.

Ok, couple of photos from the compo finals:





KK: thanks for your feedback. Completely agree about the FFT - we finished it in a hurry, should have filtered it and detect peaks. Now we can polish it a bit. He he didn't know that you were testing the pads :D. Other than that we could add a small nano-kontrols for each of the coders. DX11 compiles a lot slower that's why we used GLSL, but the problem was it still compiled slowly and this resulted in some frame drops in the end. We will just do it asynchronous next time.

DFox: we would gladly help you in setuping it, and I could come to help you organize it. Of course if every other organizer would like it. It's a pinch in the ass (not a pain thou), and the end result is very rewarding for both competitors and the audience. As KK said - it's essential to make it in parts for cooldowns. At first we needed about 15 minutes for switching from a party setup to realtime coding setup. Before the finals I just needed 5 minutes. But If you will have separate machines for the compo, and separate machines for running screens and other compos, and a mixer with lot of inputs (we had Roland v-800HD, that has just 4 inputs) then it should be smooth as silk.

so the round with four pairs takes at least 1,5 hour. You could shorten first battles to 15 minutes each, make semifinals last for 20 minutes, and finals 30 minutes. In this setup
you will need:
about 1h and 20 minutes for first round
about 50 minutes for semifinals
about 35 minutes for finals

We were lucky since we got 8 people, but if you had 16 then you would need to make some eliminations (preselection) to figure out who's the best. Also necessary is to know the competitors and pickup such pairs that will bring the best ones to the finals (so you wouldn't have the best match in the semifinals or 1st round).

ok. Except for that here are both shaders from the finals. I tried to put them into shadertoy but I failed :). If someone could try to do it then I'll be very thankful:

http://plastic-demo.nazwa.pl/bonzaj/KK_shader.glsl
http://plastic-demo.nazwa.pl/bonzaj/w23_shader.glsl

Garg: sorry for the mess I left in the upload area, I had some severe ODD problems with my data connections being regularly closed by scene.org servers. Weird!

bonzaj: thanks for the offer - i'll speak about it with the team and see if we have enough timeslots available. we could also do something like a first round on friday, semifinals on saturday and the final on sunday before the democompo or something like that.

We will actually have a dj-set with live-coding-vj stuff at the Ultimate Meeting this year. It's not the same but a bit similar. Maybe we can find some more people interested to compete at Revision then :)

True. But the UAV argument still stands for GLSL. Too bad there is no DX11 assembler in Cg compiler.

And don't tell me about "frame drops in the end" - I had my Visual completely frozen for several seconds more than few times (really upsetting event during real-time coding).


Nor anybody in the audience. :) Still, it came out as a nice turn of events.

And if you would add the nano-kontrols to the equation, I imagine you could extend the formula. E.g. when coding stops coders stand up to cheering crowd with nano-kontrols and 1-minute VJ battle starts with orgs doing the video mixing.


It's hard to do anything more than a raytraced sphere in 20 minutes, so I don't think it would be wise decision. Any coder that goes into raytracing won't have to show too much after 15 mins.

Of course, if you want to really cut the compo time and have equipment neccessary, you can set up 3 or 4 coders in one round (optimization by paralellization).

..just my coin here: don't forget about a host with the right programming kung-fu! Seeing bonzaj getting perversely excited about how the shaders expand over time really stirred up interest in the crowds :)

Kbi: regarding the Wild entries, the Willy's entry was a pure original work as well as mine was (Yans playing using hands only), but I explicitly requested it as a non-spread as I felt it was too crappy (or, at least, not enjoyable by international audience). I'm not sure about Willy's entry, but I guess he probably doesn't care ;). However, how about Leśnik's piece? The music was definitely original.

whens the ETA of the elude demo? I WANT IT ON MY PHONE! =)

argasek: => Aceman. He's the man to decide. I'm just a servant who obeys his upload instructions :)

What happened with THE futuris demo? It isnt even in THE results.txt ..

It didn't work on the compo machine, sadly.

Or well, not initially. If it did later, then I missed it.

kbi: OK. AceMan confirmed Leśnik requested to not spread his entry, too.

argasek: Dzieki za poczestowanie zacnym trunkiem!

Tomoya: cała przyjemność po mojej stronie, miło było wreszcie zobaczyć jak wyglądasz :D

Just got to a temporary home, but too tired to write a full party report. So there's a partial report:
Basically, everything was awesome, and I had great fun at the party! Huge thanks to organizers, especially for the live coding compo, it was particularly wonderful. Had a great time participating in it (I didn't expect I could write code that fast! :D). Hugs go to all competitors and especially to KK. He's very tough and very serious. I honestly think he's actually way better programmer than me and I just got lucky with my "build order" this time :D.

Anyway, I quickly fixed my shader to work on shadertoy: https://www.shadertoy.com/view/ldl3W2. Obviously, the Pad and FFT inputs are absent (but they didn't quite work as expected anyway, as KK has written above).

I guess we also need a new category and platform for pouet, as I'm confused about submitting.

I had a blast! Thanks to the organizers and all the friendly people at the party.