All Picture Lab models generate images from a text prompt. Many also support additional modes: some let you train a custom LoRA, others accept reference images for editing. The sections below show which extra capabilities each model has.
Trainable models
Trainable models generate images from a text prompt and support LoRA training — letting you teach the model your own avatar, visual style, or product. You don't need a reference image to generate; the model creates from scratch based on your prompt.
| Model | What it does |
|---|---|
| Z-Image | High-quality generation with accurate prompt following and LoRA support |
| WAN 2.2 | Photorealistic people and lifestyle images with natural faces and lighting |
Edit models
Edit models take a reference image and transform it based on your prompt — changing the scene, style, lighting, or composition while preserving identity or structure. Many of these models work best with at least one reference image. For some it is required; for others it is optional.
| Model | What it does |
|---|---|
| Qwen Edit | Prompt-based edits using reference images |
| Flux 2 Klein | Prompt-based edits using reference images |
| Qwen Image 2.0 | Fast generation and editing with balanced quality |
| Seedream 4.0 | Multi-reference control of pose, clothing, and style |
| Flux 1 Kontext Pro | Context-aware editing that preserves identity |
| Grok Imagine | Generation and editing powered by xAI Grok |
| Qwen Image 2.0 Pro | Enhanced realism and advanced text rendering |
| Nano Banana 2 | Reference-based generation for styles and drafts |
| Nano Banana Pro | High-quality generation and editing |
Generation-only models
Generation-only models create images from a text prompt with no reference required. GPT Image adds unique controls for output quality and background transparency, making it useful for product images and UI assets.
| Model | What it does |
|---|---|
| GPT Image | Text-to-image with quality and background controls |
| Flux 2 Pro | Complex scenes, detailed prompts, and final visuals |
Switch your model
Open the model selector in the top-left corner of Picture Lab. Each model shows its type badge, a short description, and a price tier. Models with a lock icon require a paid plan.
