ComfyUI Node
Lackluster Prompt Enhancer
A ComfyUI node in Lackluster/Prompt with 18 inputs and 2 outputs.
Lackluster Prompt Enhancer
- image0
- image1
- image2
- image3
- image4
- enhanced_prompt
- negative_prompt
◄prompt►
◄task_typev2v►
◄ollama_urlhttp://127.0.0.1:11434►
◄ollama_model►
◄api_formatOllama►
◄auto_enhancefalse►
◄text_input—►
◄negative_prompt►
◄temperature0.70►
◄max_tokens2048►
◄seed0►
◄prepend_system_prompttrue►
◄unload_ollamafalse►
CategoryLackluster/Prompt
Inputs (18)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | Raw instruction or generation prompt. Use the Enhance button or enable auto-enhance to rewrite with the task template. | |
| task_type | COMBO | v2v | Bernini task mode. Determines the enhancement template and system prompt used by the LLM. |
| ollama_url | STRING | http://127.0.0.1:11434 | LLM endpoint URL (Ollama or vLLM/OpenAI-compatible). |
| ollama_model | STRING | LLM model name. Leave empty to skip enhancement (pass-through). | |
| api_format | COMBO | Ollama | Ollama uses /api/chat. OpenAI/vLLM uses /v1/chat/completions. |
| auto_enhance | BOOLEAN | false | When enabled, enhances the prompt server-side on every queue. |
| image0opt | IMAGE | Reference image slot 0 = 'image0' in your prompt. Converted to base64 and sent to multimodal LLMs. | |
| image1opt | IMAGE | Reference image slot 1. Use 'image1' in your prompt. | |
| image2opt | IMAGE | Reference image slot 2. Use 'image2' in your prompt. | |
| image3opt | IMAGE | Reference image slot 3. Use 'image3' in your prompt. | |
| image4opt | IMAGE | Reference image slot 4. Use 'image4' in your prompt. | |
| text_inputopt | STRING | Text node connection input. If connected, replaces the prompt field. | |
| negative_promptopt | STRING | Custom negative prompt. Leave empty to use Bernini's default. | |
| temperatureopt | FLOAT | 0.700–2 | LLM temperature. Lower = more deterministic. GPT-5 / o-series models ignore this and use their internal default. |
| max_tokensopt | INT | 204864–8192 | Maximum tokens in the LLM response. |
| seedopt | INT | 00–2147483647 | LLM seed for reproducibility. 0 = random. Ollama only; ignored by OpenAI/vLLM endpoints. |
| prepend_system_promptopt | BOOLEAN | true | Prepend the task system prompt to the enhanced output. Useful for CLIPTextEncode / T5 conditioning. |
| unload_ollamaopt | BOOLEAN | false | Tell Ollama to unload the model from VRAM after enhancement. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| enhanced_prompt | STRING | — |
| negative_prompt | STRING | — |