Generate images - Havincy

Text to image

curl -X POST https://havincy.com/api/v1/generations \
  -H "Authorization: Bearer $HAVINCY_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "modality": "image",
    "model_id": "fal-gpt-image-2-5-sunburst",
    "prompt": "Minimalist poster of a lemon on a pastel blue background, studio light",
    "aspect_ratio": "9:16",
    "resolution": "2k",
    "count": 4,
    "wait_seconds": 45
  }'

count: 1 to 4 images in one request (each one is charged).
aspect_ratio and resolution: see the model’s options in GET /models?modality=image.

Image editing and references

Models that accept references (accepts_reference_images: true) can edit an image or use up to 4 references (product, character, style).

From a URL
From the library

{
  "modality": "image",
  "model_id": "fal-nano-banana-2",
  "prompt": "Place this bottle on a wet rock by the sea, golden hour",
  "reference_image_urls": ["https://example.com/bottle.png"]
}

Public https URLs and data:image/...;base64, URIs are accepted (JPEG, PNG, WEBP, max 10 MB). Imported references are free.

Upload once with POST /uploads, then reuse the asset id:

{
  "modality": "image",
  "model_id": "fal-nano-banana-2",
  "prompt": "Same bottle, studio packshot on white",
  "reference_asset_ids": [1234]
}

Stay on brand

Read the workspace brand kits with GET /brand_kits (tone, audience, instructions, colors, fonts) and include them in your prompts.

⌘I