---

title: caption() description: Python API reference for ciagen.caption() : auto-caption images using a vision-language model. keywords: api, caption, python, image captions

---

caption()

Generate captions for images using a vision-language model.

from ciagen import caption

caption(
    images="data/real/train/images/",
    captions_dir="data/real/train/captions/",
    engine="ollama",
    model="llava",
)

Parameters

Parameter	Type	Default	Description
images	str \| Path	required	Directory containing images to caption
captions_dir	str \| Path	required	Directory to save caption .txt files
engine	str	"openai"	Captioning engine: openai or ollama
model	str	"gpt-4o-mini"	Vision model name
api_key	str \| None	None	API key (required for OpenAI)
image_formats	list[str] \| None	None	Supported image formats

Behavior

• Creates a .txt file per image in captions_dir with the same stem name
• Skips images that already have a caption file
• Supports both OpenAI and Ollama engines