Phota Profile Picture

The "which one is this again?" node

Profile IDs are opaque strings. PhotaProfilePicture fetches a single profile's cover photo so you can see who (or what) an ID belongs to before you spend credits generating with it. It's the one-profile version of PhotaProfileGallery - think "who is this, precisely," where the gallery is "show me everybody." It's also a free API call, so it's the cheap way to sanity-check a profile before committing a billed generation to it.

How it works

The node hits the API's cover-photo endpoint for the profile you name, receives a JPEG, and converts it into a standard IMAGE tensor you can preview. The catch: a cover photo only exists once a profile is READY. Ask for a profile that's still validating, queuing, or in progress and you'll get a 404-style error - so this node is really for confirming identity on profiles that have finished training.

The inputs

• profile_id - the ID you want a face for (required).
• profile - optional typed socket from a Phota Profile Selector or Train Profile node; a connected socket overrides the typed ID, same as elsewhere in the pack.
• api_key - leave empty; use PHOTA_API_KEY or phota.ini so the key doesn't ride along in shared workflow JSON.

Output

A single IMAGE - one cover photo. Drop it into PreviewImage and you've turned an opaque string into a face.

Where it fits

A few genuinely useful spots: verifying a shared workflow's hardcoded profile_id before you run it; checking that the profile a Selector dropdown shows actually is the subject you think it is (the dropdown labels are just id [status] (tag), which is meaningless when IDs drift); or as the last sanity step before a billed generation of a client's likeness. Combined with PhotaProfileStatus, it's also a nice two-node status board: check READY, then show the face.

Install is the standard pack setup - search Phota API Nodes in ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/photalabs/comfyui-phota

…then restart. Only dependency is requests, no GPU or model files, and the fetch itself costs nothing against your credits.