Runware PhotoMaker
PhotoMaker — keep the face, change everything else
- images
- photoMaker
PhotoMaker is the identity tool from the "no training required" era: give it a few photos of a person, and it folds that face into the generation so your prompt can put them anywhere - astronaut, painter, formal portrait. It's the middle rung between "loose IP-Adapter vibe" and "train a whole LoRA," and this builder is how you wire it into a Runware model node. You feed reference images in, pick how strongly the identity holds, pick a style, and the RUNWARE_PHOTOMAKER value you get goes into a model node's photoMaker socket.
The inputs that matter
- images - required IMAGE. Your reference photos of the person. A few clear shots beat one bad one; the model is assembling an identity embedding from all of them.
- strength - 15 to 50, default 15. Read the tooltip carefully, because the direction is counterintuitive: lower values give stronger subject fidelity, higher values give more creative freedom. Default 15 is already the "hold onto the face" end. Bump it up when the prompt needs room to transform the subject (the astronaut case); keep it low when you need the face unmistakably theirs.
- style - an 11-way enum: No style, Cinematic, Disney Character, Digital Art, Photographic, Fantasy art, Neonpunk, Enhance, Comic book, Lowpoly, Line art. This is a built-in style transfer on top of the identity - basically a one-line LoRA for flavor. "No style" is the baseline; the rest are worth sampling on your own face reference because they vary wildly in how well they read.
Where it fits in identity-land
The KB's identity-preservation essay is the honest framing: the adapter/identity family peaked and then faded as instruction-editing models took over character consistency - but adapters like PhotoMaker are still the right tool for one specific job, which is getting a specific real face into a generation from a few photos with no training. That's exactly this node. What it won't do is give you full-character consistency across a whole scene (that's what a LoRA is for, and the KB is blunt that LoRAs capture far more than a face). If "keep this person's face, in this setting" is the job, PhotoMaker is the tool.
Output and wiring
Output is RUNWARE_PHOTOMAKER. It feeds a model node's photoMaker socket - only one model node in the catalog exposes one, so wire it into that and match the socket, or the node does nothing.
Install
The usual one-time pack setup:
cd ComfyUI/custom_nodes
git clone https://github.com/Runware/ComfyUI-Runware
pip install -r ComfyUI-Runware/requirements.txt
Restart, add your Runware API key (Settings → Runware API key, or RUNWARE_API_KEY).
The honest take
The one thing that trips people is the inverted strength semantics - everyone nudges strength up to get a better likeness and makes it worse. Start at 15, and only move it up if the identity is choking the prompt. And manage expectations on likeness: identity tools get you close and stop, which is why the classic power move is PhotoMaker for structure plus a final face-swap pass if you need 98%. For "I want a portrait of my friend in a studio setting," this node is most of the way there in one call.
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| images | IMAGE | — | |
| strengthopt | INT | 1515–50 | Controls the balance between preserving the subject's original features and the creative transformation specified in the prompt. Lower values provide stronger subject fidelity, higher values allow more creative freedom. |
| styleopt | COMBO | No style | Artistic style applied to the generated images. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| photoMaker | RUNWARE_PHOTOMAKER | — |