I typed one paragraph. My AI agent set up ComfyUI and Krea 2 on a rented cloud GPU and made these images on a laptop with no graphics card. Get the exact prompt, skill file and script.
Real outputs from the run in the video. Krea 2 Turbo, 1024×1024, NVIDIA L4 on Modal. Not retouched.
Your laptop doesn't need a graphics card. It needs an agent that can rent one by the second, and a spending cap so it can't surprise you.
Make a Modal account and lower the usage limit to $30, the monthly free credit. Two minutes, and it's the only part with money in it.
It installs the tools, downloads 18.6 GB of model weights once into cloud storage, starts ComfyUI on an L4 GPU and fixes its own errors.
Images land in ./outputs. The agent reports the cost, and you double-check it in the Modal bill. Mine made mistakes; the kit tells you where to look.
Everything from the video, cleaned up so your agent can skip the trial and error.
SETUP-CHECKLIST.mdThe 5 human steps, including the $30 capAGENT-PROMPT.mdThe exact paragraph + a shorter one for the kitSKILL.mdSkill file for Claude and other agentskrea2_comfy.pyModal + ComfyUI + Krea 2 Turbo scriptprompts_batch.jsonThe 14 prompts from the batch runLICENSE-NOTES.mdWhat you can do with the imagesPlain text and Python only. No installers, nothing to run blind.
This is the whole instruction. Everything after it in the video, the agent did.
My laptop has no GPU. Use Modal to run ComfyUI with the Krea 2 Turbo model on a cheap GPU. Keep the model weights in a Modal Volume so they only download once. Then generate one image from this prompt and save it to ./outputs: "a vintage film camera on a wooden desk by a window, soft morning light, real photograph, 85mm lens". When you're done, shut everything down and tell me what it cost.
Checked on 2026-09-29. Prices and limits can change.