ComfyUI Node
Bernini-R Prompt Embedding
Task-aware prompt planner: selects system prompt, encodes text with CLIP, caches embeddings.
Bernini-R Prompt Embedding
- clip
- positive
- negative
- system_prompt
- full_prompt
◄task_typeDefault / General►
◄user_prompt►
◄clip_name▾►
◄clip_typewan►
◄clip_devicecpu►
◄negative_prompt色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走►
◄delimiter
►
◄force_offloadtrue►
◄use_disk_cachetrue►
◄cache_tag►
CategoryBernini-R/Conditioning
Inputs (11)
| Name | Type | Default | Description |
|---|---|---|---|
| task_type | COMBO | Default / General | Pre-built task type. Selects a system prompt + default negative tailored to the task |
| user_prompt | STRING | Your creative prompt describing the desired video content | |
| clipopt | CLIP | CLIP text encoder. If connected, CLIP loader inputs below are ignored | |
| clip_nameopt | COMBO | CLIP model file (ignored if 'clip' input is connected) | |
| clip_typeopt | COMBO | wan | CLIP architecture type |
| clip_deviceopt | COMBO | cpu | 'cpu' saves VRAM during encoding, 'default' = GPU |
| negative_promptopt | STRING | 色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走 | Negative prompt. Leave empty to use task default |
| delimiteropt | STRING | Separator between system prompt and user prompt | |
| force_offloadopt | BOOLEAN | true | Offload CLIP to CPU after encoding to free VRAM |
| use_disk_cacheopt | BOOLEAN | true | Cache encoded embeddings to disk for reuse across sessions |
| cache_tagopt | STRING | Optional tag to isolate caches (e.g. per model or project) |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| positive | CONDITIONING | — |
| negative | CONDITIONING | — |
| system_prompt | STRING | — |
| full_prompt | STRING | — |