Видео

There are a lot of tutorials. Search for Stable Diffusion and Deforum.

@@mushroots1679 I haven't found one where it's symmetrical, it's really amazing

Just render half the screen resolution and mirror it afterwards. For the fly trough effect adjust the camera translation settings so the view does not move in or out the frame on the side you mirror it.

@@mushroots1679 thanks a lot.

Complaining on the internet about stuff you didn't do yourself also gets old imo. Then go and make trippy visuals which aren't mirrored. There are reasons my animations are mirrored and i explained it in another comment. If you are interested in a genuine discussion about bilateral symmetry in psychedelic art i'm all in for it but if you are just here to complain please go away and complain on another video.

You good?

INK❤❤ yesss

Thank you! I use Automatic1111 and the Deforum extension. Deforum does not work with the latest version of Auto1111. You have to either download Version 1.7 or use Forge (didn't tried it yet but it should work). github.com/AUTOMATIC1111/stable-diffusion-webui github.com/lllyasviel/stable-diffusion-webui-forge I also used Blender for the input Video. blender.org

@@mushroots1679 Thank you very much for your response and your precise information regarding this software. So I'm going to find out if, on the one hand, all of this is free (?), and then if it's not too difficult for me to experience. I have a very high speed Livebox and a good PC. I hope I'm well 'equipped'.. Thanks again . BBF

It's all completely free and runs on your own computer. The Checkpoint models and Loras you can download from here, also for free: civitai.com/ Most important is that you have enough VRAM. The more the better, it runs faster, you can make bigger pictures and you can do more complex renders with a lot of Controlnets. I have 10GB but wish i had more. I often have to limit my controlnets or picture size to render my animations in a reasonable time. I recommend watching tutorials on how to set everything up and how to use deforum.

@@mushroots1679 Thanks again for this new link to 'civitai'.. So it's with these 'models' and 'photos' that you make your videos? On my PC, your videos are full screen and in 16:9, but on 'civitai' I have hardly seen any models in 16:9..! With 'Blender', I think you have to put them back in 16:9..? My PC has 32.0 GB of Ram, with an Inter Core i7 processor, 3.00 GHz. I will follow your advice and watch some tutorials.. Thanks again . BBF

@bbfneroblack6663 the photos on civitai are more to show whats possible with specific loras and models. I sometimes play around with the prompts and settings from those pictures but do not use them directly. I also rarely render in an 16:9 aspect ratio because Stable diffusion is trained on a 1:1 aspect and the result looks the best near that ratio. If you look closely you will see i only render half of the screen (8:9) and add a mirror und the middle. I dindt mean RAM, i ment VRAM which is the RAM inside your GPU that can't be upgraded separately.

I didn't made the music. Just had a look and it seems this track isn't available on Spotify but others from FX23 and you can download this track for free at Ektoplazm. Download link is in the description.

This was made in Automatic1111 with Deforum. Cadence is at 2, strength at 0.8 But like i said before, those numbers probably won´t help you with anything because there are at least 10 other settings influencing the consistency. The strength is usualy the last setting i change regarding to the other settings.

No comfyui. This is made with Blender and Deforum in Auto1111. I won't share the whole script with all the settings because there are already a ton of bad and uncreative AI Animations poluting the web and i don't wanna be part of that. Try recreate it by yourself and learn what the settings do. A script with random numbers won't help you with that. If you have any specific questions regarding the settings I'm happy to help.

I made a short video in blender of moving patterns and processed it further with stable diffusion and deforum. I used the hybrid video function, some Controllnets and a lot of different Loras. If you're not familiar with those programs just look for tutorials, there are a lot. Edit: Sorry, wrong video. I didn't use an input video for this one. Only Stable Diffusion and Deforum.

Thank you! No, it's actually 1.5. SDXL takes to long on my 10GB GPU. For this one i used the "A-Zovya RPG Artist Tools" V4 model. You find it on civitai. I also used a bunch of Loras which change over time.

I hope you mean BLOWS like I do, as in 'the last thing I'd want to be listening to on a psychedelic'. I just mute this and open another YT to play what I like to go along with it, like Son of Ohm, Vintage Cucumber, Holger Czukay, Lunar Dunes, Sula Bassana or a little Nonchalant Johnson. Great visuals, though, on a lot of these, they've really upped the game.

Hey that sounds like great music you listen to! But why can't you let people enjoy what they like? I myself would never listen to this kind of music while tripping but obviously a lot of people do. I just can't wrap my head around your idea that your taste in music makes you kinda superior? I wish you a nice journey, old soul 😘

Regarding the music i agree with you but not fully with the animation. I am the creator of the animation. The "AI" i used only rendered the animation according to my settings and i had a long time getting to those. I agree that the animation is not mine anymore from the moment i uploaded it publicly to the web but not because it was made with "AI". The moment you clicked on the video it was downloaded to your computer or smartphone and you where as much an owner of the video and the data, as i am.

You would like if i made an animation with indian gods? I will try it over the weekend and let you know if i was successful. 🔱

@@mushroots1679 Awesome! Yes, I hope you can make it! I appreciate!

I wasn't able to produce a good animation with hindu deities yet. Maybe I'll try again another time.

@@mushroots1679 I appreciate for your kindness!

Thank you so much LUVAIR for all your compliments! I think it isn't really a specific workflow, i just spent a crazy amount of time trying different settings to understand how they influence each other. For most of my videos i spend at least 2 full days trying different settings until I'm happy with the results and let it render. There are at least 10 different settings that influence the smoothness and depending on the results you are trying to achieve you need slightly different settings. This one was rendered with a fixed seed thats why some details appear stationary while others move. I also used cadence 5, deleted all cadence frames and interpolated only the SD rendered frames with film x16. To delete the cadence frames i used chatGPT to write a python script that deletes the frames in a specific folder and keeps every nth frame. I also could have just used cadence 1 and a higher Z translations but that also influences the depth and 3D warping of the video a lot. I highly recommend spending a lot of time with deforum and get an understanding how all the different settings and controllnets influence the animations visually. That's also one of the reasons I'm very open about my settings but don't share the whole script because it wouldn't help anyone getting better with deforum or understand the settings.

Thanks! I really appreciate the giving some detail but not everything about the workflow. Your prompts, level of detail and smoothness is pretty unreal. I think I'm close to what I'm trying to achieve minus the sharper detail and smoothness. I'm going to try the cadence set at 1:00 and then interpolating. I was doing the interpolating inside of after Effects but it seems like stable diffusion will do a better job. 👍 Thank you. I always look forward to one of your new uploads. 🍄🤯

Then we already have the same workflow i guess. I also have an idea in the begining and then i try different things to achieve it and i think that makes the difference. Because of how AI gets hyped at the moment, a lot of people assume you just have to write a creative prompt and AI does the rest and you see it in a lot of animations. The reason i deleted a lot of settings is to force myself to try other settings. The next video is currently rendering and on it's way. I had to move my computer and couldn't render for a few days therefore the delay.

​@@mushroots1679haha Yeah that's the funny thing about AI. I know if I make this stuff and put it out there people are just going to assume that I just wrote up some texts and out came a great looking video. It's interesting you mentioned 1 cadence vs 5 cadence. Even with deleting the frames and interpolating I seem to be getting the better results with cadence 5 and delete the frames inbetween vs cadence 1 as you had said. Have you experimented with guidance images at all? I've been trying it to guide animations over time. It's a bit frustrating but interesting. Not really any tutorials on it, I think it's a sleeper. Looking forward to your next work!

Glad i could help you with the cadence settings and great to see you tried both settings. Best way to learn and understand. I tried guided frames once to add some logos to an animation. No matter what settings i tried, it somehow influenced the whole animation color wise and because the logos where mostly just white on black background, it came out bad. Even with color coherence set to a reference picture. I'm not sure if it's a bug or i missed something but i never tried it again. Maybe i come back to it with another project or idea someday, it's definitely an interesting tool to get creative.

Yes this one was with reference_adain+attn and depth. Tile, clip_vision, ip-adapter_clip or combinations of those can also help.