Skip to content

Image Generation

h-avoyan edited this page Jun 18, 2026 · 3 revisions

Image generation

65 image models spanning text-to-image, image editing, inpainting, style transfer, background tools, upscaling, and vector/SVG output.

Quick start

gen-ai generate -m flux-2-pro -p "studio shot of a ceramic cup, soft light" --ar 4:3
{ "name": "picsart_generate",
  "arguments": { "model": "flux-2-pro", "prompt": "studio shot of a ceramic cup", "aspectRatio": "4:3" } }

Providers

Provider Models Highlights
Google Nano Banana Pro, Nano Banana 2, Nano Banana Top-tier generation + natural-language editing, up to 4K
Flux Flux 2 Pro / Max / Flex, Flux Kontext Crisp, controllable; reference-guided editing
OpenAI GPT Image 2, GPT Image 1.5 Strong text rendering & instruction following
Recraft Recraft V4.1 family (20) True vector / SVG output
Ideogram Ideogram v3, Character Best-in-class typography
Seedream Seedream 5.0 Lite, 4.5 High-detail photoreal
Qwen Qwen 2 / Pro, Edit Plus Generation + editing
Luma UNI-1, UNI-1 Max Fast creative generation
Hunyuan Hunyuan V3
Reve Reve
Picsart SANA-Sprint, Image Edit, Makeup, Change/Remove BG, Enhance Picsart-native editing tools
Topaz Upscale + Enhance family (9) Upscaling & restoration

Editing shortcuts

Some tasks have dedicated, model-agnostic commands/tools that auto-pick a fitting model:

gen-ai remove-bg ./photo.jpg
gen-ai change-bg ./photo.jpg -p "a sunny beach"
gen-ai enhance ./photo.jpg
gen-ai vectorize ./logo.png
{ "name": "picsart_remove_bg", "arguments": { "imageUrls": ["https://example.com/photo.jpg"] } }

Common image parameters

Param CLI flag Notes
prompt -p Required
aspectRatio --ar Model-dependent set (e.g. 1:1, 16:9, 4:3)
resolution -r e.g. 1K, 2K, 4K (model-dependent)
count -n Number of outputs
imageUrls -i Reference / input images for editing & i2i

Exact options vary per model — check gen-ai models info <id> or picsart_model_params.

Clone this wiki locally