ComfyUI Node
Gemini Text Generation
Generate text using Gemini models.
Gemini Text Generation
- client
- response
◄prompt►
◄seed-1►
◄modelgemini-3.5-flash►
◄temperature0.70►
◄max_tokens8192►
◄top_p0.95►
◄top_k40►
◄stop_sequences►
◄response_mime_typedefault►
◄response_schema►
◄thinking_levelnone►
CategoryERPK/Gemini
Inputs (12)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | Text prompt for Gemini | |
| seed | INT | -1-1–2147483647 | Seed for reproducible outputs. Randomizes by default. |
| clientopt | GEMINI_API_CLIENT | Gemini API client from Gemini API Config node (optional if API key is configured in Settings) | |
| modelopt | COMBO | gemini-3.5-flash | Gemini model to use for generation |
| temperatureopt | FLOAT | 0.700–2 | Creativity level (0.0=focused, 2.0=very creative) |
| max_tokensopt | INT | 8192256–65536 | Maximum length of response |
| top_popt | FLOAT | 0.950–1 | Nucleus sampling - cumulative probability threshold (0.0=disabled) |
| top_kopt | INT | 40-1–100 | Top-k sampling - limit token selection (0=disabled) |
| stop_sequencesopt | STRING | Stop generation at these sequences (one per line, leave empty to disable) | |
| response_mime_typeopt | COMBO | default | Output format (use application/json for JSON mode) |
| response_schemaopt | STRING | JSON schema for structured output (only used with application/json, leave empty for free-form JSON) | |
| thinking_levelopt | COMBO | none | Reasoning depth. Works on Gemini 2.5 and 3.x; the node translates to thinking_budget (2.5) or thinking_level enum (3.x) automatically. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| response | STRING | — |