Put two characters in one image

tendre.red can place two trained characters in the same scene while keeping both faces recognizable, using the 2-character scene mode in LoRA generation. It needs two separate, already-trained LoRA, one per character; the mode does not work with InstantID-only characters that have no LoRA.

Train both characters first

Each character needs its own LoRA, trained the normal way from its own photo dataset. Do not try to combine two people into one LoRA, they must be trained and stored separately for a duo to work.

Turn on the 2-character scene

In LoRA generation, switch on 2-character scene and pick a LoRA for the left side and one for the right side.

Describe each side, and the shared scene

Fill in a short description for the left character and one for the right (outfit, hair, anything that tells them apart), plus the shared scene in the main prompt (add "two people" and describe the setting). A landscape format works best for two figures side by side.

Balance the result

If one character comes through stronger, lower its LoRA strength slightly and raise the other's. The space balance slider moves the divide between the two sides if one figure needs more room in the frame.

When it struggles

Very close framing or overlapping poses are harder to keep separate. If the ComfyCouple node used for this mode is missing, restart the app, it installs with tendre.red.