Coil the Bass! | ||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
![]() |
|
|||||||||||||
|
popularity : 43% |
|||||||||||||
alltime top: #37642 |
|
|||||||||||||
added on the 2025-02-17 15:27:18 by teadrinker ![]() |
popularity helper
comments
I don't the M8, but this one was one of the most amazing entries imho, the sound vibes and concept.
rulez added on the 2025-02-17 15:40:49 by Optimus 

that was cool ♥ had the stream audio running on a quite massive sound system and it really payed off ♪♫
🍵🎶
Very interesting thing ;)
Best of Lovebyte for me! What kind of synth model is it ?
Hehe only teadrinker makes that bass possible; easy guess on the stream ;)
Interesting concept and visuals
nice
thank you everyone!
>What kind of synth model is it ?
I'm using the MicroW8 snd(int index) hook, which returns a single float sample (alternating left and right channels, so it's called 44100x2 times per second)
The sound generation is running the same code as is driving the visuals, it's based on the plasmosis plasma ( https://www.reddit.com/r/tinycode/comments/1aq13hj/text_plasmosis_14_byte_plasma_winner_of_16b/ ) running with a buffer of 960x21 (and at a speed synced to match buffer size * 60 fps), buffer data is played back at half speed (compared to the buffer id) so at a different speed than the simulation. If frames are dropped due to performance issues, it will run the simulation multiple times on the visual side to remain in sync.
In audio-dsp terms, it's kind of like a single delay line with a 6db IIR filter combined with some kind of strange 2d blur, and value offset / wrap-around.
>What kind of synth model is it ?
I'm using the MicroW8 snd(int index) hook, which returns a single float sample (alternating left and right channels, so it's called 44100x2 times per second)
The sound generation is running the same code as is driving the visuals, it's based on the plasmosis plasma ( https://www.reddit.com/r/tinycode/comments/1aq13hj/text_plasmosis_14_byte_plasma_winner_of_16b/ ) running with a buffer of 960x21 (and at a speed synced to match buffer size * 60 fps), buffer data is played back at half speed (compared to the buffer id) so at a different speed than the simulation. If frames are dropped due to performance issues, it will run the simulation multiple times on the visual side to remain in sync.
In audio-dsp terms, it's kind of like a single delay line with a 6db IIR filter combined with some kind of strange 2d blur, and value offset / wrap-around.
base and visualizer in perfect harmony
very artsy! love it! one of my faves from the compo!
submit changes
if this prod is a fake, some info is false or the download link is broken,
do not post about it in the comments, it will get lost.
instead, click here !