---

title: generate() description: Python API reference for ciagen.generate() : generate synthetic images from real images using Stable Diffusion + ControlNet. keywords: api, generate, python

---

generate()

Generate synthetic images from real images using Stable Diffusion + ControlNet.

from ciagen import generate

result = generate(
    source="data/real/train/images/",
    output="data/generated/",
    extractor="canny",
    sd_model="fennecinspace/sd-v15",
    cn_model="lllyasviel/sd-controlnet-canny",
    num_per_image=3,
    prompt="a person walking in a park",
    seed=42,
    device="cuda",
)

Parameters

Note

All parameters except source, output, extractor, sd_model, and cn_model are optional.

Parameter	Type	Default	Description
source	str \| Path	required	Directory containing real source images
output	str \| Path	required	Directory to save generated images
extractor	str	required	Condition extractor: canny, openpose, segmentation, mediapipe_face
sd_model	str	required	HuggingFace model ID for Stable Diffusion
cn_model	str	required	HuggingFace model ID for ControlNet
num_per_image	int	1	Number of synthetic images per real image
seed	int \| list[int]	34567	Random seed for reproducibility
device	str	"cuda"	"cuda" or "cpu"
prompt	str \| list[str] \| None	None	Positive prompt(s)
negative_prompt	str \| None	None	Negative prompt
quality	int	30	Number of inference steps
guidance_scale	float	7.0	Classifier-free guidance scale
use_captions	bool	False	Use per-image caption files as prompts
captions_dir	str \| None	None	Directory with caption .txt files
modify_captions	bool	False	Vary captions using vocabulary substitution
vocabulary_template	str \| None	None	Vocabulary config name for caption modification
generation_size	int	10	Number of prompt variations to generate
cn_extra_settings	dict \| None	None	Extra kwargs for ControlNet loading
image_formats	list[str] \| None	["png","jpg","jpeg"]	Supported image formats

Returns

dict with:

Key	Type	Description
total_generated	int	Number of images successfully generated
output_path	str	Absolute path to the output directory
source_images	int	Number of source images processed
errors	list[tuple]	(image_path, error_message) for failures