TouchDesigner Tutorial - Generate AI Images with Stable Diffusion using Image-to-Image Generation

Hey! In this tutorial, we'll go over how to use Stable Diffusion with a custom component I created to generate images in TouchDesigner. The project supports 2 forms of input using prompt generation and image to image so you can use any TOP in TouchDesigner as a starting point.

Huge thank you to Peter Whidden for his support on this and his work on computerender which makes this project possible

0:00 Overview and Examples
2:00 Setup The Computerender Component
2:36 Generating Single Images
4:04 Generate Multiple Images
6:18 Working with Iterations
8:05 Working with Guidance
9:23 Customize Resolution
10:19: Handling Errors
11:08 Auto Saving Images
12:18 Prompt Engineering with Lexica
13:09 Ethics of Copying Prompts
13:50 Generating More Image Examples
15:22 Using Any TOP as Input

[anti-alias]

Amazing.. cannot wait until realtime video generation is possible

[ParkinsonMax]

Thank you for setting this up, it's really awesome. Fantastic work!

[EmptyDreamTV]

Thanks I was watching old TDNeuron videos all yesterday and this is pretty much what I was actually looking for so Thank you so much 😊

[brvoigt]

At 11:30, to make it write to the correct folder after selecting a location using the File Browser, edit the script in generate_file_name to add the slash between folder and filename
op('moviefileout1').par.file = parent().par.Imagefolderpath+ '/' + parent().par.Prompt...

[vjkuru]

amazing!. Is there a way to run the stable diffusion model locally without the API?

thanks, greetings from Mexico

[irenearias]

thank you much. you are a literal angel and saved my life and i swear i'll join your patreon when i'm not a broke college student 💗💗💗💗

[mahshadfaridfar]

Amazing. Thank you for sharing this information.💯

[RusticRaver]

amazing, tx for that, it's been 6 month I was waiting for AI to hit touchdesigner with good explanations! I'd also be interested by the python side of things.

[archbramzia]

Thank you for share this great project with us ! And I have a problem at first, which is i clicked and nothing happened, I checked Textport and DATs, it shows Failed to establish a new connection. Then I figured out this problem, by using Astrill VPN stealth mode. After that everything works properly:) This suggestion specially for China users.

[vickiwu]

thank you so much for setting this up! It is super helpful and great work!

[prismatic.visuals]

Awesome work - and very well explained!

[flupke]

any way to implment this on local gpu with local stable diffusion install ?

[ajk_]

This is incredible! thank you so much.

[若林徐]

What do I do if there is a error: numpy.core.mutiarray failed to import?

[alexeyparkhomenko]

fantastic, thanks for sharing your work!

[grapessyrop]

i dont why but its not working for me. it has a giant red X o the computerrender. i have the 2023 touchdesigner and it seems to be the latest version. so what is wrong?

[zacharykrevitt]

This is amazing!! is there a way to bring in your own models?

[suleymanyilmaz]

Hey Torin, thanks for amazing tutorial. Is it possible to use other models inside the computerender module? And which version of SD we are using with compute render ? Again thank you thank you .

[thebusinessofsade]

This was an amazing video. I am trying to get serious about doing this type of stuff and I was wondering how you learned to code python specifically for touchdesigner. I am thinking of learning python but want to learn it in a way that is directly relevant to touchdesigner. Also how did you find computerrender are there other websites like it?

[filippoboi]

Hi Torin and thank you very very much! I have a question, I tried to use my webcam as a img input and it's very intresting the mix with the prompt, I would like to have a video output, even if it's not perfect and in a different framerate, is it possible using your computerender component?