Hot starts, batch inference, and what's next for Runpod Serverless. Webinar June 25.

Prompt Scheduling with Disco Diffusion on Runpod

This guide introduces prompt scheduling in Disco Diffusion, a technique that lets you shift prompts dynamically throughout an image generation run. Learn.

Prompt Scheduling with Disco Diffusion on Runpod

You could schedule advanced stuff like cuts before, but now you can schedule PROMPTS in Disco Diffusion! That means that you can alter the prompt over the course of your diffusion steps to achieve some really awesome results. Read more to find out how!

DiscoArt, the platform that Runpod's Disco Diffusion is built on, just introduced prompt scheduling today. If you need to learn how to start a pod with the GUI, you can check out this Disco Diffusion on Runpod article.

All you need right now to schedule prompts is to switch on the "Enable Advanced Custom Settings" under the "Custom Settings" expander. Then you can paste in your JSON that contains your prompt schedule like so:

Disco Diffusion advanced custom settings JSON with scheduled text prompts, weights, and clip guidance

Here's the starting prompt used here as an example if you want to copy and edit it:

Now you can totally do stuff like introduce new prompts half-way through your render. You can even turn clip guidance for each prompt on and off independently. I can't wait to try this out and post some results.

Note that using the advanced overrides will ignore the prompts you put into the normal UI. We will be building an advanced prompt editor soon to make this experience better. <3

Author profile: Zhen Lu

Related articles

View All
Deploy When Available is now GA

Deploy When Available is now GA

Queue for any GPU spec, even one that's fully rented out, and we'll deploy it the moment capacity opens up. No more refreshing the console or running a sniping tool.

All
The Chips Got Faster. The Stack Didn't.

The Chips Got Faster. The Stack Didn't.

Explore why faster chips have shifted the bottleneck to AI infrastructure, and what that means for teams running production workloads.

All

Build what’s next.

Build, train, and scale AI workloads on Runpod with cloud GPUs, Serverless, and Clusters.