User acreil explained how to solve your delay problem (without using the z~ object) in another thread:
http://puredata.hurleur.com/sujet-7057-feedback-algorithm

You want to make subpatch with block size 1. In that patch, you create an array of length 1 and then write/read to that array. This has to be in a separate patch because dac~ and adc~ require block size 64.

Regarding your second question, I do not know what you are describing as the "play object". Maybe attach a patch.

I can't say for sure without seeing your patch, but this sounds like a problem I solve with a liberal sprinkling of pix_separator all over the place.

Thanks.

Here it is. Kind of a messy patch: I use two screens. There are way too many oscillators. It sounds crazy because the block size is set up for visuals and ruins the audio. My favorite idea was using video effects in a feedback chain on the audio. So pix_convolve is messing with the oscillators.

I find that when the oscillators are very close to an exact multiple of the sample-rate divided by the block size (or something like that, idk), it looks & sounds cool. I have yet to try new waveforms with filters. I think that would be cool to modulate a filter on a square-wave that's tuned to the block-size.

http://www.pdpatchrepo.info/hurleur/oscillators_final_nopresets.pd

I hate to echo the last response... but screen capture wins.
I never got pix_record to work on linux - didn't even try on windows.
With screen capture, you get audio AND video.

Try Microsoft's free screencap software.
http://www.microsoft.com/expression/products/Encoder4_Overview.aspx

Here's an example of some harsh sig2pix I did a few months ago. Expression encoder did a great job at 15 fps.

http://www-ui.is.s.u-tokyo.ac.jp/~takeo/teddy/teddy.htm
Wow dude! Just found this website from the lowend3d forum.
This is an amazing java applet that allows you to sketch and export .obj files extremely easily.
Blender is the bomb for serious stuff, but this program gives you instant gratification and that total psycho outsider aesthetic.
edit: this for windows. i tested one object and it works.

Hi.

I have been playing with sig2pix for the past week, but hadn't thought of mixing in images before. I got some good results tonight after seeing your post. It's definitely distorted.

You need one gemhead for pix_image and one for pix_sig2pix~. Pix_texture comes after you combine the two images.

Some good ways to combine are diff, chroma_key, composite, multiply, and add.

The images have to be the same size to work. I like to use pix_crop because you can scroll around a large image while retaining the correct dimensions.

I also like to add in pix_pix2sig~ before the pix_texture. You output each channel to a send~ or a delay, then you can add that back in with your R,G,B,A oscillators. This way you can have one color effect another, or possibly create feedback.

I find that the feedback gets quickly distorted. I don't know how to emulate my friend's analog RGB synth where he gets insane feedback fractals. To calculate 768x576 pixels at 30 fps you would need a sample rate of 13 million hertz. D'oh.