Generation
These nodes create images, video, audio, text, or 3D models from your inputs. A socket is an input or output point on a node.
Generate Image
Generate or edit an image with the selected model.
Inputs: Text and reference images, as supported by the selected model.
Outputs: One generated image.
Settings
| Setting | Effect |
|---|---|
| Provider and model | Choose a model from the model browser. |
| Aspect ratio and resolution | Available values depend on the model. |
| Search | Google Search on direct Nano Banana Pro and Nano Banana 2. Image Search on Nano Banana 2. |
| Model settings | Open the controls card for settings supplied by the selected model. |
| Fallback | An alternative model and its own settings for an automatic retry. |
Tips
- Use the model tables to check provider-specific settings.
- Use the node history to compare earlier results.
Generate Video
Generate, edit, or extend video with a selected model.
Inputs: Text, images, video, or audio through the model-specific sockets.
Outputs: One video.
Settings
| Setting | Effect |
|---|---|
| Provider and model | Select a video model to expose its sockets. |
| Model settings | Duration, aspect ratio, resolution, and other fields depend on the model. |
| Fallback | An alternative model and settings for a retry. |
Tips
- For Seedance 2 I2V, use First/Last Frame or Reference Images, not both modes together.
- Seedance reference limits are nine images, three videos, and three audio inputs.
Generate Audio
Generate speech, sound effects, or other audio from a selected model.
Inputs: Text by default, with extra sockets where the model defines them.
Outputs: One audio result.
Settings
| Setting | Effect |
|---|---|
| Provider and model | Choose an audio model from Kie.ai, fal.ai, Replicate, or another discovered provider. |
| Model settings | Voice, format, and other fields depend on the model. |
| Fallback | An alternative model and settings for a retry. |
Tips
- Use Array batch mode for a list of speech prompts.
- Send the result to Output or Video Stitch.
LLM Generate
Use a large language model (LLM) to generate text or describe an image.
Inputs: Text and optional images.
Outputs: Generated text.
Settings
| Setting | Effect |
|---|---|
| Provider and model | Google, OpenAI, or Anthropic. See the current catalog. |
| Temperature | Controls variation. Range 0–1 for Anthropic, or 0–2 for Google and OpenAI. |
| Max tokens | Limits output length from 256 to 16,384 tokens. Tokens are units of text used by the model. |
| Fallback | An alternative model and settings for a retry. |
Tips
- Send generated lists to Array for batch prompts.
- A retired saved model ID triggers a replacement note.
Generate 3D
Generate a 3D model from text or an image.
Inputs: Text or images through model-specific sockets.
Outputs: A 3D model.
Settings
| Setting | Effect |
|---|---|
| Provider and model | Select a compatible Replicate, fal.ai, or WaveSpeed model. |
| Model settings | Fields depend on the selected model. |
| Fallback | An alternative model and settings for a retry. |
Tips
- Connect the 3D output to 3D Viewer to inspect it and capture an image.
- Generate 3D does not use Array batch mode.
ComfyUI App
Run an imported ComfyUI workflow as one node.
Inputs: Typed sockets selected from the imported workflow. An empty node has no sockets.
Outputs: Selected workflow results, including image, text, video, audio, or 3D data.
Settings
| Setting | Effect |
|---|---|
| Workflow | Attach an editor save, API export, Blueprint, or saved node. |
| Exposed controls | Choose inputs, settings, and outputs in the import or edit dialog. |
| Backend | Set Comfy Cloud, This computer, or Remote in Settings → ComfyUI. |
Tips
- Use saved nodes to reuse an attached workflow and its values.
- The ComfyUI guide covers setup, formats, previews, and the curve editor.
Next steps
Continue to Image Processing.