Nodes/ComfyUI-kaola-qwen-image2.1-prompt-rewrite/Qwen 2.1 Prompt Enhancer Rewrite
ComfyUI Node

Qwen 2.1 Prompt Enhancer Rewrite

A ComfyUI node in Qwen2.1 with 19 inputs and 4 outputs.

By kana112233·Created 3 days ago·Updated 2 days ago· 2
Qwen 2.1 Prompt Enhancer Rewrite
  • qwen_pe_model
  • image_1
  • image_2
  • image_3
  • image_4
  • image_5
  • image_6
  • image_7
  • image_8
  • image_9
  • positive_prompt
  • wh_ratio
  • ratio_follow
  • thinking
◄taskt2i►
◄system_prompt_file▾►
◄prompt—►
◄temperature1.00►
◄top_p0.95►
◄top_k20►
◄presence_penalty1.50►
◄max_new_tokens16256►
◄seed42►
CategoryQwen2.1

Inputs (19)

NameTypeDefaultDescription
qwen_pe_modelQWEN_PE_MODELConnect the loaded model from Qwen2.1 PE Loader.
taskCOMBOt2iTask type: 't2i' for text-to-image prompt expansion, 'edit' for image-based rewriting.
system_prompt_fileCOMBOSystem prompt template file located in ComfyUI/models/LLM/prompts/.
promptSTRINGEnter the short prompt or instruction you want to enhance.
temperatureFLOAT1.000–2Controls randomness. Lower values make output more deterministic, higher values increase creativity.
top_pFLOAT0.950–1Nucleus sampling threshold. Restricts the vocabulary selection range.
top_kINT200–100Limits vocabulary to the top K most likely tokens.
presence_penaltyFLOAT1.500–2Penalizes repeated words. Higher values prevent the model from getting stuck in loops.
max_new_tokensINT162561–32768Max generation length. Keep this large because the model outputs a long <think> reasoning process.
seedINT420–18446744073709550000Random seed. Fixed seeds produce identical deterministic results.
image_1optIMAGE—
image_2optIMAGE—
image_3optIMAGE—
image_4optIMAGE—
image_5optIMAGE—
image_6optIMAGE—
image_7optIMAGE—
image_8optIMAGE—
image_9optIMAGE—

Outputs (4)

NameTypeDescription
positive_promptSTRING—
wh_ratioSTRING—
ratio_followSTRING—
thinkingSTRING—