Start with what you want to make, then pick a workflow. I keep the website updated, so check the linked pages for the latest files and setup steps.
Make a still image
- For natural-looking portraits, including characters you already have a LoRA for, start with Krea 2 Turbo.
- For quick prompt tests, use Z Image Turbo.
- To change an image while keeping a character, outfit or product as a reference, try Qwen Image 2.1 reference editing.
The Image Generation Runpod template is the starting workspace for these image workflows. Check which model downloads you need before deploying.
Make a video with sound
- Have a scene in mind? Try MiniMax H3 text to video. Have a starting image? Use MiniMax H3 image to video to carry that look into the clip.
- Want an LTX workflow with a second upscale pass? Start with LTX 2.5 image to video or LTX 2.5 text to video.
Use the MiniMax H3 template or LTX template for the model you choose.
Control the movement
To put your character into an existing driving video, use Wan Animate or SCAIL 2 character animation. Both start with a character image and a video whose motion you want to follow.
If you want to animate one image instead, try Wan 2.2 image to video. The Wan Runpod template covers the Wan family.
Before you deploy
Choose a workflow before renting a GPU. The memory you need changes with the model file, precision, resolution and other settings. Check the template requirements and current Runpod price, then enable only the models you need. Run ComfyUI on Runpod walks through the setup.
If two workflows sound right, run a short test with the same input in each. Keep the one that gives you the result you want at a cost and speed that work for you.