Precise Reference
info
This feature was previously called "Character Reference" and was renamed to "Precise Reference" in February 2026.
Control character appearance and art style with reference images. Precise Reference allows you to maintain consistent character designs and apply specific art styles across generations.
Features
- Multiple reference images: Combine multiple references to apply different characters and styles
- Reference types:
"character": Reference character appearance only"style": Reference art style only"character&style": Reference both character and style (default)
- Fine control: Adjust fidelity and strength for each reference
Basic Example
from novelai.types import CharacterReference, GenerateImageParams
# Single character reference
character_references = [
CharacterReference(
image="reference.png", # Base64 string or file path
type="character", # "character", "style", or "character&style"
fidelity=1.0, # Reference fidelity (0.0 to 1.0, default: 1.0)
strength=1.0, # Reference weight (0.0 to 1.0, default: 1.0)
)
]
params = GenerateImageParams(
prompt="1girl, standing in a garden",
model="nai-diffusion-4-5-full",
character_references=character_references,
)
# Execute (assuming client is initialized)
# images = client.image.generate(params)
Advanced: Multiple References
Combine character and style references from different images:
character_references = [
CharacterReference(
image="character.png",
type="character", # Character appearance only
fidelity=1.0,
strength=0.75,
),
CharacterReference(
image="style.png",
type="style", # Art style only
fidelity=1.0,
strength=0.75,
),
]
params = GenerateImageParams(
prompt="1girl, standing, rating:general, very aesthetic",
model="nai-diffusion-4-5-full",
character_references=character_references,
)
Parameters
image(required): Reference image (file path, Base64 string, or PIL Image)type: Reference type"character": Apply character appearance only"style": Apply art style only"character&style": Apply both (default)
fidelity: How closely to match the reference (0.0-1.0, default: 1.0)strength: Relative weight when using multiple references (0.0-1.0, default: 1.0)