Portrait Master 2.9.2 (Legacy)

This is the one that started it. Before the pack split into five modular nodes, there was one giant PortraitMaster node with roughly seventy inputs, and it was everywhere in 2023 and 2024 - the reddit launch threads for it hit +90 upvotes and people were building entire SDXL portrait workflows around it. In 3.x the author split it into Base Character, Skin Details, Style & Pose, Make-up, and Face Generator, but kept this monolith around under the "Old versions" category so every workflow and PNG that references it keeps working.

So when is it the right tool? When you already have an old workflow that uses it, or when you genuinely want everything in one panel. Otherwise reach for the split nodes - a single node with fifty inputs and fifty random_* toggles is a lot to look at, and the modular versions are easier to reason about. The one thing the legacy node has that the others don't: it outputs two strings, a ready positive prompt and a ready negative prompt, so it's closer to a drop-in prompt generator than the modular chain.

How it works

Same mechanism as the rest of the pack - dropdowns and sliders become weighted tag text, lowercased, joined with commas. What's different:

• RETURN_NAMES are positive and negative. The positive is the full assembled portrait prompt. The negative starts from your negative_prompt field and, when photorealism_improvement is set to enable, automatically appends (shinny skin, shiny skin, reflections on the skin, skin reflections:1.35) - the classic SDXL "don't give me wax" guard.
• prompt_start / prompt_additional / prompt_end let you wrap your own text around the generated core. prompt_start defaults to raw photo, (realistic:1.5), which is a solid baseline for SDXL photorealism.
• Per-field random_* booleans replace the random 🎲 dropdown entries of the new nodes - flip random_hair_color and it rerolls each run.
• It prints the assembled prompt to the ComfyUI console on every run. Annoying if you run long batches, useful the first few times to see exactly what it's building.

One legacy-specific gotcha: the photorealism_improvement input here is an enum (enable/disable), not the boolean toggle the new nodes use.

The inputs that matter

The defaults are already set to something sensible (Close-up, Man, 30, Afghan, Beefy, Amused, Agfa Vista…) - you can queue it the second it loads and get a usable portrait prompt. From there, the high-leverage edits are shot, gender, age, hair_style, facial_expression, and negative_prompt. Everything else is refinement. Set a random_* boolean to true and it'll surprise you; the pack's randomization is genuinely good for exploring.

Installing it

It's the same pack - install comfyui-portrait-master once and every node including this legacy one shows up:

cd ComfyUI/custom_nodes
git clone https://github.com/florestefano1975/comfyui-portrait-master

Then restart ComfyUI. Or search "Portrait Master" (florestefano1975) in ComfyUI Manager. No dependencies, no model downloads.

Common issues

• Loading an old workflow scrambles the values. This is the defining legacy-node problem - the community has reported dropdowns holding values that don't exist in the current lists (a "Red" face shape, a weight sitting in a hair-style field) after loading PNGs saved against older versions. "Creating monsters," as one user put it. Re-set the fields manually.
• Full body still isn't full body. The long-running complaint. The author's own advice: reduce the face/skin detail weights, describe the clothing explicitly, and put (close up:1.5) in the negative prompt.
• The prompt is enormous. Stacking every category at high weights produces a monster string that dilutes attention on SDXL-lineage CLIP encoders. Set the fields you don't care about to -.
• The console spam. It prints both prompts every run. That's by design, not a leak.