Generators
Kaisei-Fukaya edited this page Sep 23, 2022
·
9 revisions
Generator nodes produce new assets by interpreting inputs.
The Image from Text node produces bitmap images based on provided text prompts.
- Different machine-learning models can be selected from the "Models" dropdown.
- For example models may be similar to DALL·E 2 or Stable Diffusion
- The default resolution is 512x512 pixels, however you can provide your own resolution values via the "Resolution Width" and "Resolution Height" inputs.
Graph view | Inspector view |
---|---|
The Mesh from Photo node produces textured meshes from photographic images.
- The "Input Type" can be switched between "Single-view" and "Multi-view".
- When set to "Single-view" a single photo input is required.
- When set to "Multi-view" four photo inputs are required, with each representing a subject at different viewing angles. This results in higher quality outputs.
- Different machine-learning models can be selected from the "Models" dropdown.
- These models specialise in specific types of asset so it is important to pick the correct one for the task.
- By default this node will provide one output per input, however you can specify how many results to generate per input via the "Amount to generate" port.
Graph view | Inspector view |
---|---|
The Mesh from Photo node produces meshes from hand-drawn images.
- The "Input Type" can be switched between "Single-view" and "Multi-view".
- When set to "Single-view" a single sketch input is required.
- When set to "Multi-view" four sketch inputs are required, with each representing a subject at different viewing angles. This results in higher quality outputs.
- Different machine-learning models can be selected from the "Models" dropdown.
- These models specialise in specific types of asset so it is important to pick the correct one for the task.
- By default this node will provide one output per input, however you can specify how many results to generate per input via the "Amount to generate" port.
Graph view | Inspector view |
---|---|
The Interpolator 3D node generates a new textured mesh based on two provided textured meshes.
- Different machine-learning models can be selected from the "Models" dropdown.
- These models specialise in specific types of asset so it is important to pick the correct one for the task.
- A slider determines the similarity of the output asset to either input. By default this is in the middle, meaning that outputs will have "equal" similarity to both inputs.
Graph view | Inspector view |
---|---|
The Style Transfer 2D node generates an image by combining the content and style of two different bitmap inputs.
- Different machine-learning models can be selected from the "Models" dropdown.
- For example models may be similar to this.
Graph view | Inspector view |
---|---|
The Style Transfer 3D node generates a textured mesh by combining the content and style of two different textured mesh inputs.
- Different machine-learning models can be selected from the "Models" dropdown.
- For example models may be similar to this.
Graph view | Inspector view |
---|---|