// GENERATIVE AI
Image, Video and LLM Generation Nodes
Generative nodes sit in the same graph as the processing nodes. LLM__CHAT writes or analyses text, IMAGE__GENERATE makes images, and VIDEO__GENERATE makes video, so one pipeline can go from a prompt to a finished asset.
Nodes used
LLM__CHATDeepSeek, Llama, Qwen or Grok, with strength and effort sliders; cheapest first.
IMAGE__GENERATEFlux 2 or SDXL Turbo on DeepInfra, or Grok Imagine, with a strength slider.
VIDEO__GENERATEGrok Imagine video, DeepInfra models (Wan, Pixverse, Pruna, Veo), GPU Wan/LTX, or a Grok storyboard.
How to set it up
- 01Add LLM__CHAT to draft or refine a prompt, and connect its output to IMAGE__GENERATE.
- 02Send the image on to IMAGE__UPSCALE or VIDEO__GENERATE.
- 03Publish and pass your own prompt as an input when you call the endpoint.
Costs
Runs use Quarks. Every account gets 100 free Quarks each month. Generative nodes are usually the most expensive part of a graph, so test with the cheaper model settings first.