Viltrum

Documentation

Documentation

Back to website

Build

Search

Search

⌘K

⌘K

Build

AI Generation Nodes

Combine Model, Positive and Negative prompt nodes with the Image Generator to produce consistent, on-brand results.

7 min read

AI generation in Viltrum is built from small, composable nodes. You choose a model, describe what you want, describe what to avoid, and pass all three into a generator node that does the work.

The building blocks

  • Model selects the checkpoint to use, for example DreamShaper 6 (SD1.5).

  • Positive holds the prompt describing what you want to see.

  • Negative lists what to keep out of the result, such as text, extra objects or busy backgrounds.

  • Image Generator takes the model and both prompts, then returns one or more images.

Image Generator settings

  • Randomness sets the seed. Use a fixed number to reproduce a result exactly.

  • Control mode switches between Fixed and Random seeds for each run.

  • Quality steps controls how many refinement steps the model takes. More steps are slower but sharper.

  • Prompt strength decides how closely the image follows your prompt, from loose to literal.

  • Sampling method picks the sampler. dpm++ 2M is a reliable default.

Writing better prompts

Lead with the subject, then style, then lighting and composition. Keep negative prompts short and specific. If results drift between runs, lock the seed with Control mode set to Fixed.

Click Generate on the Positive node to preview a single image before running the full workflow. It is the fastest way to tune a prompt.

Was this page helpful?

Yes

No

Build

Search

⌘K

Build

AI Generation Nodes

Combine Model, Positive and Negative prompt nodes with the Image Generator to produce consistent, on-brand results.

7 min read

AI generation in Viltrum is built from small, composable nodes. You choose a model, describe what you want, describe what to avoid, and pass all three into a generator node that does the work.

The building blocks

  • Model selects the checkpoint to use, for example DreamShaper 6 (SD1.5).

  • Positive holds the prompt describing what you want to see.

  • Negative lists what to keep out of the result, such as text, extra objects or busy backgrounds.

  • Image Generator takes the model and both prompts, then returns one or more images.

Image Generator settings

  • Randomness sets the seed. Use a fixed number to reproduce a result exactly.

  • Control mode switches between Fixed and Random seeds for each run.

  • Quality steps controls how many refinement steps the model takes. More steps are slower but sharper.

  • Prompt strength decides how closely the image follows your prompt, from loose to literal.

  • Sampling method picks the sampler. dpm++ 2M is a reliable default.

Writing better prompts

Lead with the subject, then style, then lighting and composition. Keep negative prompts short and specific. If results drift between runs, lock the seed with Control mode set to Fixed.

Click Generate on the Positive node to preview a single image before running the full workflow. It is the fastest way to tune a prompt.

Was this page helpful?

Yes

No

Create a free website with Framer, the website builder loved by startups, designers and agencies.