Image models
Image models
Magnific gives you access to 30+ AI image generation models, each optimized for different needs in style, speed, and output quality.
Not sure where to start? Select Auto in the AI Image Generator and the AI will pick the best model for your prompt.
How to select a model
Open the AI Image Generator
Go to the Image Generator from the AI Suite homepage.
Click the Model selector
Find it in the settings panel next to the prompt field.
Browse or search
Browse by provider or type a model name in the search bar.
Select your model
Choose a model and adjust aspect ratio, references, and other settings.
Click Generate
Your image will be ready in seconds.
The same prompt can look very different depending on the model. If you're not happy with a result, try switching — especially when moving between photorealistic and artistic styles.
Model comparison table
Model | Description | Best for |
|---|---|---|
Auto | AI picks the best model for your prompt automatically. | General use, when you are unsure which model to choose |
Cinematic | Studio-grade photorealism with cinematic composition. 2K-4K output. | Editorial imagery, product campaigns, storytelling scenes |
Mystic 2.5 | Versatile with good prompt adherence. | General use, high-quality results with minimal adjustments |
Mystic 2.5 Flexible | Best prompt adherence in the Mystic family. | Character design, branding, projects needing precision |
Mystic 2.5 Fluid | Better quality with faster generation. | Visual campaigns and large image sets needing speed and quality |
High quality with detailed textures. | Portraits, rich textures, poster-like illustrations | |
Classic | Solid, reliable model. | General generation when you want a good result without fuss |
Classic Fast | High quality, very fast and low cost. | Quick conceptual visuals, rapid iterations |
Latest Google model. Fast with strong multi-reference support. 2K-4K. | Brand consistency, product campaigns, multi-image sets | |
Sharper accuracy and robust character preservation. 2K-4K. | Branded visuals, expressive characters, precise rendering | |
Google Nano Banana 2 Lite | Lighter, faster version of Nano Banana 2. | Quick drafts, high-volume iteration, cost-sensitive generation |
Google Nano Banana | Fast, one-shot accuracy, consistency across outputs. | Fashion try-ons, product swaps, multi-reference generation |
Google Imagen 4 Ultra | Highest-quality Imagen model. Exceptional text rendering. | Advertising campaigns, editorial, professional product visuals |
Google Imagen 4 | Great photorealism and prompt adherence. | Product mockups, marketing visuals with precise typography |
Google Imagen 4 Fast | Speed-optimized version of Imagen 4. | Quick iterations, idea testing, efficient workflows |
Google Imagen 3 | Ultra high quality with rich detail and refined atmosphere. | Architecture, interiors, detailed scenes |
Flux.2 Max | Highest-quality Flux model. Up to 8 references, real-time data grounding. | Product marketing, e-commerce, precise color control |
Structure, consistency, stable composition. Up to 4 references. | Product lines, branded assets, character-based campaigns | |
Flux.2 Flex | Expressive detail with broader prompt interpretation. | Concept art, atmospheric scenes, stylized illustrations |
Flux.2 Klein | Lightweight and efficient. Fast with solid prompt adherence. | Quick drafts, high-volume generation, everyday creative tasks |
Flux.1 Kontext Max | High-precision editing with fine control over text and layout. | Text-heavy visuals, typography, visual identity workflows |
Flux.1 Kontext Pro | Style and object consistency across prompts. Supports real-time editing. | Characters, branded visuals, product consistency |
Flux.1.1 | Focused on pure precision. | Complex scenes, concept art, prompt-specific compositions |
Flux.1 | Versatile with good prompt adherence and personalization. | General-purpose generation, realistic scenes, illustrations |
Flux.1 Realism | Optimized for ultra-realistic images. | Product renders, portraits, fashion, architecture |
Flux.1 Fast | Balances prompt adherence and efficiency. | Idea exploration, quick variations, social media content |
GPT 1.5 High | Detailed rendering, controlled variation, strong structural consistency. | Branding, campaigns, product imagery, art-directed visuals |
GPT 1.5 | Stylistic continuity across variations via multi-reference blending. | Social media, marketing drafts, general-purpose imagery |
GPT 1 HQ | Higher quality outputs with better detail and prompt fidelity. | Marketing materials, detailed illustrations, professional visuals |
GPT | General-purpose, versatile and easy to use. | Quick concepts, exploratory designs, fast generation |
Pixel-perfect text rendering, native 4K output, and photorealistic results with natural language editing and character. | Product shots, editorial imagery, campaigns, editorial series, brand imagery | |
GPT 2 Mini | Faster, lighter GPT 2 variant with extended quality tiers (low → max). | Quick iterations needing GPT 2's text accuracy at lower cost |
Krea 2 | Photoreal generation with medium/high quality tiers. | General photorealistic generation, quick concepting |
Luma Uni-1.1 | High-fidelity generalist model with mid/max quality tiers. | Versatile scenes, character and product work needing extra polish |
Magnific One | Magnific's flagship in-house model. Up to 4K, medium/high quality tiers. | Premium campaigns, high-res branded visuals, flagship-quality work |
MAI Image 2.5 | Microsoft's multimodal image model with broad reference support (style, character, product, locations). | Experimental workflows needing varied reference types |
Seedream 5 PRO | Fast with strong multi-reference support. | Brand consistency, product campaigns, multi-image sets |
Production-level consistency. Up to 14 references, multi-round editing. | Campaigns, character storytelling, social media content sets | |
Strong character control and multi-image fusion. Up to 4K. | Photorealistic characters, product visuals, ads and e-commerce | |
Seedream 4 | Text-to-image, image editing, and multi-image generation. | Storyboarding, branding, editorial, ultra-HD projects |
Seedream 4 4K | Same Seedream 4 pipeline, native 4K output. | Ultra-HD projects, large-format print, detailed campaigns |
Ideogram | Best-in-class for graphic design and typography. | Posters, logos, cards, any design with readable text |
Ideogram 3 with Image Reference | Ideogram 3 plus character consistency via image references. | Text and character-consistent branded visuals |
Ideogram 4 | Turbo/mid/high quality tiers, sharper typography. | Text and character-consistent branded visuals, premium design work |
P-image Ideogram | Ideogram × Pruna AI family across four quality tiers (very-low to high). Strong text rendering and product/character fidelity, up to 2K. | Fast drafts to polished outputs in one family, text-heavy designs, product and portrait shots |
Built for graphic design. Structured compositions, typography-friendly. | Branding, campaigns, posters, editorial layouts | |
Recraft V4 Pro | 2K version of Recraft V4 with enhanced precision and visual judgment. | High-res campaigns, premium branding, detailed typography |
Recraft V4.1 | Refined structure and typography precision. | Branding, posters, design-grade layouts needing max Recraft fidelity |
Runway | Image generation with up to 3 visual references. | Visual storytelling with recurring characters or settings |
Experimental and expressive. Bold, unconventional compositions. | Creative projects where visual risk is intentional | |
Qwen Image 3.0 | Next-gen Qwen model, refined multilingual text and layout accuracy. | Text-intensive and multilingual visuals for early testers |
Z-Image | Ultra-realistic at impressive speed. Bilingual text accuracy. | Product imagery, editorial scenes, rapid creative iterations |
For full details on how credits and unlimited generations work, see AI credits and limits or Plans and pricing.