Nodes/ERPK Collection/Claude Prompt Enhancer
ComfyUI Node

Claude Prompt Enhancer

Turn 'a cat' into a shootable scene

By eRepublik-Labs·Created 12 months ago·Updated 4 days ago· 2
Claude Prompt Enhancer
  • client
  • enhanced_prompt
◄prompt►
◄stylephotorealistic►
◄detail_leveldetailed►
◄seed-1►
◄temperature0.70►
◄max_tokens1024►
◄use_streamingfalse►
◄model(inherit from client)►
◄effort(model default)►

Claude Prompt Enhancer is the fastest win in the ERPK Claude section: type "a cat," pick a style, and get back a long, rich, image-model-ready prompt that reads like a photography brief. It exists because the gap between what humans type and what diffusion models need is enormous - a bare subject produces a bare image, and this node spends a Claude call to fill in the lighting, lens, composition, and mood for you.

The mechanism is about as direct as it gets. Each of the 51 styles is a distinct system prompt that tells Claude what kind of expert it should be - "You are an expert photography director. Include: lighting setup, camera settings, lens choice, depth of field…" for photorealistic, a cinematography expert for cinematic, a fantasy art director for fantasy_art, and so on through cyberpunk, anime, oil_painting, watercolor, concept_art, minimalist, maximalist, and more. Your prompt goes in, the style's system prompt shapes the output, and you get a single enhanced prompt STRING out the other side. It's prompt engineering as a dropdown.

The inputs

  • prompt - the simple thing you want enhanced (e.g. "a cat")
  • style - the dropdown of 51 styles, default photorealistic
  • detail_level - how much detail to add
  • seed - cache control; randomize for fresh results each run, fixed to reuse
  • temperature - 0 to 1, default 0.7
  • max_tokens - default 1024, up to 4096
  • client - optional, falls through to ERPK Settings
  • use_streaming - boolean, off by default; the tooltip honestly notes streaming may not display in real time in ComfyUI

The single output is enhanced_prompt (STRING). Wire it into the prompt input of whatever image node you're feeding - a local checkpoint's CLIP text encode, or any of this pack's hosted image nodes. The natural chain is Claude Prompt Enhancer → image node, so your terse idea arrives as a well-described scene.

Install

It's part of the ERPK Collection:

cd ComfyUI/custom_nodes
git clone https://github.com/eRepublik-Labs/comfyui-nodes-erpk.git erpk
cd erpk && pip install -r requirements.txt

Requires anthropic>=0.96.0 and an Anthropic API key in ERPK Settings (right-click canvas). Or "ERPK Custom Nodes" via ComfyUI Manager, restart, done.

Where it fits - and where it doesn't

Honestly, this node is a great on-ramp if you're new to prompting: it teaches you, by example, what a good prompt contains. The output is usually verbose - sometimes too verbose for local checkpoints, where a paragraph of camera jargon can dilute a style rather than sharpen it. If you're feeding a local SDXL or Flux pipeline, you may want detail_level low and to keep the style focused. For hosted models that reward long natural-language descriptions, go detailed. And a practical note: because each run is a paid Claude call, the seed widget exists to let you cache and reuse an enhancement you like instead of re-billing every queue. Randomized seed means fresh (and slightly different) text every run - if you want the same enriched prompt back, fix it.

CategoryERPK/Claude

Inputs (10)

NameTypeDefaultDescription
promptSTRINGSimple prompt to enhance (e.g., 'a cat')
styleCOMBOphotorealisticEnhancement style to apply
detail_levelCOMBOdetailedHow much detail to add to the prompt
seedINT-1-1–2147483647Seed for cache control. Randomizes by default to ensure fresh results each run.
clientoptCLAUDE_API_CLIENTClaude API client (optional if API key is configured in Settings)
temperatureoptFLOAT0.700–1Ignored: current Claude models reject temperature (kept so saved workflows load).
max_tokensoptINT1024256–128000Maximum length of enhanced prompt. Thinking counts toward this limit. Current Claude models allow up to 128K.
use_streamingoptBOOLEANfalseEnable streaming (may not display in real-time in ComfyUI)
modeloptCOMBO(inherit from client)Override the client's model for this call. Without a client the default is the Claude API client's default model.
effortoptCOMBO(model default)How much effort Claude spends (thinking depth, tool calls, answer length). Higher costs more tokens and time. (model default) leaves the model's own default.

Outputs (1)

NameTypeDescription
enhanced_promptSTRING—