ComfyUI Node
AuK Prompt Enhance X
Local Prompt Enhancer driven by Qwen2.5-Omni-3B. It uses an optional ASR transcript to render the model instruction, calculate task-aware duration, and prepare whisper audio levels. Wire all applicable outputs forward.
AuK Prompt Enhance X
- encoder
- audio
- instruction
- seconds
- task
- prepared_audio
◄instructionmake her sound excited and say welcome home►
◄context►
◄max_new_tokens256►
CategoryWorkflowX/Audio/AuK
Inputs (5)
| Name | Type | Default | Description |
|---|---|---|---|
| encoder | WORKFLOWX_AUK_ENCODER | Encoder from AuK Encoder Loader. Its checkpoint must include the language head - re-run tools/convert.py --component encoder on the original Qwen directory if this errors. | |
| instruction | STRING | make her sound excited and say welcome home | Loose request in any wording. The language model picks the closest AuK task and fills its template. |
| contextopt | STRING | Optional extra context, e.g. the STRING output of AuK Whisper Transcribe on the source audio. | |
| audioopt | AUDIO | Source/reference audio. Required for edit-task duration calculation and whisper audio preparation. Connect the prepared_audio output to Instruction Encode.audio. | |
| max_new_tokensopt | INT | 25632–1024 | Generation budget for the language-model answer. |
Outputs (4)
| Name | Type | Description |
|---|---|---|
| instruction | STRING | — |
| seconds | FLOAT | — |
| task | STRING | — |
| prepared_audio | AUDIO | — |