Multi-Character Positioning
Assign specific prompts and positions to each character.
When generating multiple characters, their traits (hair color, eye color, clothes) can sometimes blend together. Multi-Character Positioning lets you assign specific prompts and positions to each character, preventing this issue.
from novelai.types import Character, GenerateImageParams
characters = [
Character(
prompt="1girl, red hair, blue eyes, school uniform",
enabled=True,
position=(0.3, 0.5), # Left (X: 0.3, Y: 0.5), same as "B3"
),
Character(
prompt="1boy, black hair, green eyes, casual clothes",
enabled=True,
position=(0.7, 0.5), # Right (X: 0.7, Y: 0.5), same as "D3"
),
]
params = GenerateImageParams(
# General prompt can be used, but details should be in characters list
prompt="two people standing together, holding hands, best quality",
model="nai-diffusion-4-5-full",
size=(832, 1216), # Size must be explicitly specified
characters=characters,
)Positions are only enforced when at least one character sets position.
Leave it out on every character to let the AI place them (the web UI's "AI's Choice").
If only some characters have a position, the others are centered.
position accepts (x, y) in the 0.0–1.0 range or a 5x5 grid preset from "A1" to "E5" ("C3" is the center).
On V4/V4.5 the coordinates are snapped to the center of their 5x5 grid cell (0.1, 0.3, 0.5, 0.7, 0.9) before sending, like the web UI does, so (0.2, 0.5) is sent as (0.3, 0.5).
V5 supports free-form positioning and sends the coordinates as given.