I have been taking a break, country side, not getting coverage most of times, and no laptop even. But clean air and lots of stars away from light pollution
I’m kinda missing all you guys, and from what I can catch up you’ve all been busy
Ok, here’s a totally inappropriate joke: women are like pools. They require too much maintenance for the time we spend inside them.
And another (this one strikes close to home): women are like hurricanes. When they come they’re hot and moist. When they leave they take your house and your car.
I don’t even know what I’m doing but I’ll be here all week
Divorced? Yeah, I did the same in '94 after being married only 4 years. Only good thing that came out of that relationship was my two boys (Who are now grown men). Current GF has been sticking around and putting up with my BS for twenty years now. I can't seem to run her off no matter what I do lol. We never mention the "M word"..
Thanks, I've still got that one working, trying to dial in the right type of music to go with it.
For that one, I did attempt a 3D forest for the scene module, but the initial results were a bunch of cylinders and cones which looked nothing like a forest. I'm sure I could talk it into making it more detailed, but for now, I'm liking the results I'm getting in 2D mode.
2D is probably best for the effect you have there. I tried 3D as well but like you it’s mostly sphere trees. There’s a recurring joke “imagine a spherical cow”, well in this case imagine a spherical tree 😂
When playing around with multiple textures on 3d objects, it very quickly occurred to me that very often in Visual for spatial parameters, what is needed is continuous 360 degree parameters.
So, for example, when you want to rotate a 3d shape and have it go round and round rather than go round and jump back to the beginning to start again. Is there a way to do that without coding? In my example, I just made do with going back and forth motion to avoid jumps. Can knobs for exposed parameters be continuous 360 degree knobs in some way? If not, could such a knob mode be added as an option @giku?
Gemini added a parameter to auto rotate (which in retrospect I’m not sure why I didn’t use in the example posted!?) so I realize it can be coded in but it sort of feels like that shouldn’t have to be necessary for a control such as rotate which by its very nature is continuous.
I can’t remember whether your current 2d experiments have used it but you could ask your AI coder to explore adding a parallax control to introduce greater sense of depth? It would need some camera movement to become apparent but could be interesting.
Indeed, I do have some discussions going on with gpt regarding just that. With my MIDI forest project though, I was trying to combine two different scenes, each with different species of trees, while using chroma keys to help in the layering.
Oh, yes, I see. Thanks @pedro. Which leads me to another question: is it not possible to make/use your own 3d shaders?
I’m probably completely missing the concept (for a change!) but if I try to use the 3d camera with the shader Gemini wrote yesterday it doesn’t connect up?
I would have thought that the shader containing a 3d shape meant it was counted as a 3d shader but it seems not? Then I noticed that there’s no option to import or write your own code into the 3d shader module hence my question above.
Is it possible to use the 3d camera or xyz transform with the shader written yesterday?
You can prompt the full shader code, including the ray marcher and everything else, to render in 3D directly inside the Shader module. It becomes a black box, but technically it can do everything what 3D Shader does.
Or, you can take a more modular approach and prompt only a 3D Mask and XYZ Transform, then use them as modular components inside a 3D Shader rack. In this approach you can reuse, combine solids, try with different cameras, raymarchers etc..
So the 3D Shader is like a shader extension, that manages raymarching and the signal flow in a modular way.
Modularity is the main difference: just decide if you want LLM to generate everything also visualization or rather separate, reusable components.
----
If you made a complete 3d visualization in Gemini you can't use it in the 3D Shader rack (it's a black box, not modular), but you can ask LLM to make a separate 3D Masks, Transforms, Raymarchers and Camera presets out of it.
Okay, many thanks @giku, I think I sort of understand the differences in approach you’re describing.
So, the reason the shader from yesterday can’t be used with 3d camera is because it contains all of the elements within itself, including its own camera elements, and is a self contained unit (that’s what you mean by black box I’m assuming?) - correct?
What I still don’t understand though is what the different modular approach actually is that I could have taken to achieve the same results? I don’t understand what I would have needed to ask Gemini to code (if anything) which could then be used with a 3d camera? A 3d Mask with texture inputs (if that’s even possible?)? Or, in this case, was a self contained shader module the best route to go?
And if you ask AI to generate a 3d mask and xyz transform, how do you import them - what does the AI actually generate? A preset file?
Apologies for my lack of understanding, I do realize how frustrating it can be when others don’t understand concepts which are straightforward to you.
so @Robin you can write a whole shader that does everything, it will basically render all pixels, it will output a color for each, and is what people normally call a pixel shader as you see on shadertoy. You can ask an LLM to write you code for that, provided you give it instructions to adhere to the convention that the shader expects. Just include a shader module, expand it and change to code, and you’ll see the bare bones basic structure it has to have. Or just ask the LLM “convert this GLSL shader to MSL” then paste the code inside the module and see how it goes
Then there’s the modular approach giku designed, the shader kit stuff, where things are deconstructed into modular bits, and where you need a 3d raymarcher, a 3d camera, 3d masks or transfoms. If you go that route with LLMs you’ll need to instruct it much more precisely as each of these components has specific syntax. For instance a mask has a “Mask” variable that has to be calculated, or a color module has a “Color”
This modular approach is more versatile but also more fragmented, since there are more “moving parts”. It’s not a black box as you can see the different parts moving
I would be glad to help you further (to the best of my knowledge) but it’s a lot of typing right now, so it would have to wait
You can copy paste text. Indeed there are no instructions available yet on how to generate 3D Mask code, the best way is to point AI to the Drambo Visual manual (it will read it) then copy paste some existing 3D Mask presets and ask if it's ready to create it's own presets. Then prompt and copy / paste to Drambo.
The second way (very powerful and instant)
Use Drambo on Mac
Install a CLI , or AI client for vibe coding
Assuming it's CGPT, open CGPT app, switch to Codex (I'm not sure if it's on the free plan, should be)
Create a project and set it's folder to Drambo visual presets folder
Ask it to scan and understand existing presets, give it a link to the manual as well
The guy is an expert now, you can ask him to create drmodule preset files and they are instantly available in Drambo :)
IMO Gemini is not the best choice, CGPT (Codex) or Antrophic models are way better.
----
Inception: you can paste this conversation to CGPT and ask it for further instructions :)
Comments
lololololol....
I have been taking a break, country side, not getting coverage most of times, and no laptop even. But clean air and lots of stars away from light pollution
I’m kinda missing all you guys, and from what I can catch up you’ve all been busy
Oh wait, what’s that video even about? Not that I dislike the cover concept
Lololol 🤣 I just woke the GF with my burst of laughter and now I'm in big trouble 😅
Love me some sister sledge though
It's about the risk that all creatives make, the 9 to 5 office or taking the plunge and becoming a fully fledged Artist. :)
Just tell her one of your internet crazy friends made a joke she wouldn’t understand. That should soothe things up.
see why I’m divorced?
I'm not going to laugh.
Nope, I didn't even chuckle.
👏🏾👏🏾👏🏾👏🏾👏🏾👏🏾👏🏾👏🏾👏🏾
Ok, here’s a totally inappropriate joke: women are like pools. They require too much maintenance for the time we spend inside them.
And another (this one strikes close to home): women are like hurricanes. When they come they’re hot and moist. When they leave they take your house and your car.
I don’t even know what I’m doing but I’ll be here all week
Divorced? Yeah, I did the same in '94 after being married only 4 years. Only good thing that came out of that relationship was my two boys (Who are now grown men). Current GF has been sticking around and putting up with my BS for twenty years now. I can't seem to run her off no matter what I do lol. We never mention the "M word"..
I’m creative, I like to think so, it’s the 9 to 5 thing that eats my guts
Btw, I'm working on a cassette "Walkman" version scene.
I have to go work tomorrow. Expletive 🤬
Shoot, I have to go to work today! Looks at time - not yet, I have 33 minutes
33’ should give you plenty of time to post another one of your masterpieces (btw, loved your midi forest)
Thanks, I've still got that one working, trying to dial in the right type of music to go with it.
For that one, I did attempt a 3D forest for the scene module, but the initial results were a bunch of cylinders and cones which looked nothing like a forest. I'm sure I could talk it into making it more detailed, but for now, I'm liking the results I'm getting in 2D mode.
2D is probably best for the effect you have there. I tried 3D as well but like you it’s mostly sphere trees. There’s a recurring joke “imagine a spherical cow”, well in this case imagine a spherical tree 😂
My drambo logo bouncing through the trees was one of my proudest achievements. Not inigo level but I still had a lot of fun with it
By the way, I forgot to ask.
When playing around with multiple textures on 3d objects, it very quickly occurred to me that very often in Visual for spatial parameters, what is needed is continuous 360 degree parameters.
So, for example, when you want to rotate a 3d shape and have it go round and round rather than go round and jump back to the beginning to start again. Is there a way to do that without coding? In my example, I just made do with going back and forth motion to avoid jumps. Can knobs for exposed parameters be continuous 360 degree knobs in some way? If not, could such a knob mode be added as an option @giku?
Gemini added a parameter to auto rotate (which in retrospect I’m not sure why I didn’t use in the example posted!?) so I realize it can be coded in but it sort of feels like that shouldn’t have to be necessary for a control such as rotate which by its very nature is continuous.
There’s multiple ways you can get there. The xyz transform for starters. If you’re raymarching then the camera itself has useful presets
I can’t remember whether your current 2d experiments have used it but you could ask your AI coder to explore adding a parallax control to introduce greater sense of depth? It would need some camera movement to become apparent but could be interesting.
Indeed, I do have some discussions going on with gpt regarding just that. With my MIDI forest project though, I was trying to combine two different scenes, each with different species of trees, while using chroma keys to help in the layering.
btw. i think the scene needs an ability to generate MIDI events
Agreed, that would actually be very cool.
One step to closer to Arduino, animatronics and live interactive art installations . :)
Oh, yes, I see. Thanks @pedro. Which leads me to another question: is it not possible to make/use your own 3d shaders?
I’m probably completely missing the concept (for a change!) but if I try to use the 3d camera with the shader Gemini wrote yesterday it doesn’t connect up?
I would have thought that the shader containing a 3d shape meant it was counted as a 3d shader but it seems not? Then I noticed that there’s no option to import or write your own code into the 3d shader module hence my question above.
Is it possible to use the 3d camera or xyz transform with the shader written yesterday?
@giku
You can prompt the full shader code, including the ray marcher and everything else, to render in 3D directly inside the Shader module. It becomes a black box, but technically it can do everything what 3D Shader does.
Or, you can take a more modular approach and prompt only a 3D Mask and XYZ Transform, then use them as modular components inside a 3D Shader rack. In this approach you can reuse, combine solids, try with different cameras, raymarchers etc..
So the 3D Shader is like a shader extension, that manages raymarching and the signal flow in a modular way.
Modularity is the main difference: just decide if you want LLM to generate everything also visualization or rather separate, reusable components.
----
If you made a complete 3d visualization in Gemini you can't use it in the 3D Shader rack (it's a black box, not modular), but you can ask LLM to make a separate 3D Masks, Transforms, Raymarchers and Camera presets out of it.
Okay, many thanks @giku, I think I sort of understand the differences in approach you’re describing.
So, the reason the shader from yesterday can’t be used with 3d camera is because it contains all of the elements within itself, including its own camera elements, and is a self contained unit (that’s what you mean by black box I’m assuming?) - correct?
What I still don’t understand though is what the different modular approach actually is that I could have taken to achieve the same results? I don’t understand what I would have needed to ask Gemini to code (if anything) which could then be used with a 3d camera? A 3d Mask with texture inputs (if that’s even possible?)? Or, in this case, was a self contained shader module the best route to go?
And if you ask AI to generate a 3d mask and xyz transform, how do you import them - what does the AI actually generate? A preset file?
Apologies for my lack of understanding, I do realize how frustrating it can be when others don’t understand concepts which are straightforward to you.
so @Robin you can write a whole shader that does everything, it will basically render all pixels, it will output a color for each, and is what people normally call a pixel shader as you see on shadertoy. You can ask an LLM to write you code for that, provided you give it instructions to adhere to the convention that the shader expects. Just include a shader module, expand it and change to code, and you’ll see the bare bones basic structure it has to have. Or just ask the LLM “convert this GLSL shader to MSL” then paste the code inside the module and see how it goes
Then there’s the modular approach giku designed, the shader kit stuff, where things are deconstructed into modular bits, and where you need a 3d raymarcher, a 3d camera, 3d masks or transfoms. If you go that route with LLMs you’ll need to instruct it much more precisely as each of these components has specific syntax. For instance a mask has a “Mask” variable that has to be calculated, or a color module has a “Color”
This modular approach is more versatile but also more fragmented, since there are more “moving parts”. It’s not a black box as you can see the different parts moving
I would be glad to help you further (to the best of my knowledge) but it’s a lot of typing right now, so it would have to wait
@Robin
No worries, it's my role to explain that.
The first paragraph: correct
Second: AI can generate both text and preset.
You can copy paste text. Indeed there are no instructions available yet on how to generate 3D Mask code, the best way is to point AI to the Drambo Visual manual (it will read it) then copy paste some existing 3D Mask presets and ask if it's ready to create it's own presets. Then prompt and copy / paste to Drambo.
The second way (very powerful and instant)
IMO Gemini is not the best choice, CGPT (Codex) or Antrophic models are way better.
----
Inception: you can paste this conversation to CGPT and ask it for further instructions :)
----
We are doomed :D