Pollinations Image Generation

Generate images from text descriptions using the Pollinations.ai API.

When to use

• User asks to generate, create, draw, or render an image
• User mentions Pollinations, or any of the supported model names

How to use

Run the generation script at {baseDir}/scripts/generate.mjs with node. Pass all parameters as --key value CLI flags.

Required parameters

Parameter	Description
--prompt	Text description of the image to generate. You should craft a detailed, high-quality prompt based on the user's request.
--model	AI model to use (see model list below)

Optional parameters

Parameter	Description	Default
--output	Output file path. Parent directories will be created automatically. If omitted, saves to current directory with a timestamped filename	auto
--width	Image width in pixels	1024
--height	Image height in pixels	1024
--seed	Random seed for reproducible results. Use -1 for random	0
--enhance	Set to true to let AI improve the prompt for better results	false
--negative_prompt	Things to avoid in the image	"worst quality, blurry"
--safe	Set to true to enable safety content filters	false
--quality	Image quality: low, medium, high, hd (gptimage only)	medium
--image	Reference image URL(s), comma/pipe separated for multiple	-
--transparent	Set to true for transparent background (gptimage only)	false

Available models

Free models (always available):

• flux — Fast, high-quality image generation (default)
• zimage — High-quality image generation
• imagen-4 — Google Imagen 4
• klein — Klein image generation
• klein-large — Klein large image generation
• gptimage — GPT Image (supports quality and transparent options)

Paid models (requires POLLINATIONS_PAID=true):

• seedream / seedream-pro — Seedream image generation
• kontext — Kontext model
• nanobanana / nanobanana-pro — NanoBanana image generation
• gptimage-large — GPT Image large variant

Examples

Basic: just prompt and model

node {baseDir}/scripts/generate.mjs --prompt "a beautiful sunset over mountains" --model flux

Custom size (--width, --height)

node {baseDir}/scripts/generate.mjs --prompt "a panoramic futuristic cityscape at night" --model imagen-4 --width 1536 --height 768

Reproducible result (--seed)

node {baseDir}/scripts/generate.mjs --prompt "a red panda sitting on a branch" --model klein --seed 42

AI-enhanced prompt (--enhance)

node {baseDir}/scripts/generate.mjs --prompt "cat" --model zimage --enhance true

Negative prompt (--negative_prompt)

node {baseDir}/scripts/generate.mjs --prompt "a professional product photo of sneakers" --model flux --negative_prompt "blurry, low quality, watermark, text"

Safety filter (--safe)

node {baseDir}/scripts/generate.mjs --prompt "a fantasy battle scene" --model klein-large --safe true

Quality control (--quality, gptimage only)

node {baseDir}/scripts/generate.mjs --prompt "an oil painting of a library interior" --model gptimage --quality hd

Transparent background (--transparent, gptimage only)

node {baseDir}/scripts/generate.mjs --prompt "a cartoon robot mascot" --model gptimage --transparent true

Reference image (--image)

node {baseDir}/scripts/generate.mjs --prompt "make it look like a watercolor painting" --model kontext --image "https://example.com/photo.jpg"

Save to a specific path (--output)

node {baseDir}/scripts/generate.mjs --prompt "a watercolor landscape" --model flux --output /tmp/landscape.jpg

Combining multiple options

node {baseDir}/scripts/generate.mjs --prompt "a detailed fantasy map" --model gptimage --width 2048 --height 2048 --quality hd --seed 123 --negative_prompt "blurry, text" --safe true --output ./map.png

Important notes

• If --output is specified, the file is saved to that path; otherwise a timestamped filename is used in the current directory.
• Model availability depends on the POLLINATIONS_PAID env var. If POLLINATIONS_PAID is not true, only free models can be used.
• Choose the model based on the user's requirements. Default to flux for images if no preference.
• When crafting prompts, be detailed and descriptive for best results.
• Always tell the user the output file path after generation completes.