> For the complete documentation index, see [llms.txt](https://docs.flora.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.flora.ai/models/image-models.md).

# Image Models

Here are all the image models currently available in FLORA.

## Generation Models

These models create new images from text prompts.

<table><thead><tr><th width="220">Model</th><th width="120">Est. time (s)</th><th>Description</th><th>Best for</th></tr></thead><tbody><tr><td><strong>Ad Delayer</strong></td><td>105</td><td>Split a flat ad into editable background, product, and copy layers.</td><td>Split a flat ad into editable background, product, and copy layers.</td></tr><tr><td><strong>Flux 2</strong></td><td>30</td><td>High-resolution output, great for print.</td><td>Fast, customizable generation with acceleration options</td></tr><tr><td><strong>Flux 2 Flex</strong></td><td>45</td><td>Adjustable quality/speed balance for production.</td><td>Consistent outputs across generations, reliable results</td></tr><tr><td><strong>Flux 2 Klein 4B</strong></td><td>25</td><td>Precise modifications using natural language descriptions.</td><td>Targeted edits, natural language-driven image changes</td></tr><tr><td><strong>Flux 2 Klein 9B</strong></td><td>50</td><td>Enhanced realism, crisper text and better color adherence.</td><td>Realistic outputs with accurate text and color rendering</td></tr><tr><td><strong>Flux 2 LORA</strong></td><td>35</td><td>Flux 2 image editing with LoRA style support</td><td>Flux 2 image editing with LoRA style support</td></tr><tr><td><strong>Flux 2 Lora Trainer</strong></td><td>5700</td><td>Flux 2 LoRA trainer ($0.0064/step, priced at the 1000-step default)</td><td>Flux 2 LoRA trainer ($0.0064/step, priced at the 1000-step default)</td></tr><tr><td><strong>Flux 2 Max</strong></td><td>60</td><td>Exceptional realism, precision, and consistency.</td><td>Premium photoreal outputs, maximum quality generation</td></tr><tr><td><strong>Flux 2 Pro</strong></td><td>45</td><td>Top-tier Flux quality with multi-reference support.</td><td>Professional-grade generation, best BFL quality</td></tr><tr><td><strong>Flux 2 Turbo</strong></td><td>25</td><td>Speed-optimized Flux 2 model.</td><td>Fast iteration, quick previews, time-sensitive workflows</td></tr><tr><td><strong>FLUX 3 Image</strong></td><td>140</td><td>Precise edits that leave other pixels untouched, layout by bounding box, up to 10 references, and 4K output.</td><td>Precise edits that leave other pixels untouched, layout by bounding box, up to 10 references, and 4K output.</td></tr><tr><td><strong>FLUX 3 Image Klein</strong></td><td>10</td><td>Fast FLUX 3 generation and editing in seconds, with up to 10 references, seeds, and 4K output.</td><td>Fast FLUX 3 generation and editing in seconds, with up to 10 references, seeds, and 4K output.</td></tr><tr><td><strong>Flux Dev</strong></td><td>20</td><td>Quick drafts and rapid iteration.</td><td>Rapid prototyping, mood-boards, style fusion, abstract compositions</td></tr><tr><td><strong>Flux Pro 1.1</strong></td><td>24</td><td>Ultra-high resolution for large-format work.</td><td>Professional-grade generation, premium quality outputs</td></tr><tr><td><strong>GPT Image</strong></td><td>80</td><td>Great prompt comprehension, natural language edits.</td><td>Narrative-driven visuals, branding mock-ups, precise in-painting &#x26; style transfer</td></tr><tr><td><strong>GPT Image 1.5</strong></td><td>75</td><td>High-fidelity images with strong prompt adherence.</td><td>Premium quality generation, transparent backgrounds, precise control</td></tr><tr><td><strong>GPT Image 2</strong></td><td>220</td><td>Flagship image model from OpenAI.</td><td>High-fidelity image generation, prompt-accurate visuals, and production-ready assets.</td></tr><tr><td><strong>GPT Image 2.5 Flare</strong></td><td>25</td><td>Flagship image model from OpenAI, built for speed.</td><td>Flagship image model from OpenAI, built for speed.</td></tr><tr><td><strong>GPT Image 2.5 Sunburst</strong></td><td>35</td><td>Flagship image model from OpenAI, tuned for detailed work.</td><td>Flagship image model from OpenAI, tuned for detailed work.</td></tr><tr><td><strong>Grok Imagine</strong></td><td>20</td><td>Fast image editing model from xAI.</td><td>Quick generation with xAI's latest image capabilities</td></tr><tr><td><strong>Grok Imagine Quality</strong></td><td>30</td><td>Quality image editing model from xAI.</td><td>Quality image editing model from xAI.</td></tr><tr><td><strong>Ideogram 3.0</strong></td><td>18</td><td>Best for logos and art with crisp, readable type.</td><td>Text-in-image rendering, stylized aesthetics with curated presets</td></tr><tr><td><strong>Ideogram 4.0</strong></td><td>18</td><td>Producing crisp visuals with accurate text rendering.</td><td>Producing crisp visuals with accurate text rendering.</td></tr><tr><td><strong>Ideogram 4.5</strong></td><td>20</td><td>Crisp typography and structured edits from sketches, depth and references.</td><td>Crisp typography and structured edits from sketches, depth and references.</td></tr><tr><td><strong>Kling O1</strong></td><td>65</td><td>Precise edits with text, image, or video guidance.</td><td>Detailed image editing with high precision</td></tr><tr><td><strong>Krea 2 Large</strong></td><td>30</td><td>High-fidelity, stylized image model.</td><td>High-fidelity, stylized image model.</td></tr><tr><td><strong>Krea 2 References Large</strong></td><td>30</td><td>Generate stylized results using reference images as input.</td><td>Generate stylized results using reference images as input.</td></tr><tr><td><strong>Krea 2 References Medium</strong></td><td>30</td><td>Generate stylized results using reference images as input.</td><td>Generate stylized results using reference images as input.</td></tr><tr><td><strong>Krea 2 Turbo LORA</strong></td><td>30</td><td>Stylized image model with LORA support.</td><td>Stylized image model with LORA support.</td></tr><tr><td><strong>Krea 3</strong></td><td>10</td><td>Fast image generation and editing with optional thinking, up to 8 references.</td><td>Fast image generation and editing with optional thinking, up to 8 references.</td></tr><tr><td><strong>Lora Trainer</strong></td><td>120</td><td></td><td></td></tr><tr><td><strong>Luma Photon</strong></td><td>35</td><td>Realistic lighting and rendering for objects and architecture.</td><td>Versatile generation across diverse styles, fast iteration</td></tr><tr><td><strong>Nano Banana</strong></td><td>50</td><td>Fast and lightweight, good for iterating.</td><td>Fast, intuitive photo edits &#x26; multi-image blends for social-ready transformations</td></tr><tr><td><strong>Nano Banana 2</strong></td><td>80</td><td>Google's new state-of-the-art image model.</td><td>Next-gen image quality with Google's latest capabilities</td></tr><tr><td><strong>Nano Banana 2 Lite</strong></td><td>20</td><td>Google's new state-of-the-art image model.</td><td>Google's new state-of-the-art image model.</td></tr><tr><td><strong>Nano Banana 2.1</strong></td><td>40</td><td>Google's fast image model with adjustable thinking, successor to Nano Banana 2.</td><td>Google's fast image model with adjustable thinking, successor to Nano Banana 2.</td></tr><tr><td><strong>Nano Banana Pro</strong></td><td>90</td><td>Highest quality from Google, up to 4K.</td><td>Premium 4K generation, professional multi-image composition, highest quality outputs</td></tr><tr><td><strong>Patina</strong></td><td>45</td><td>Generate 3D PBR material maps from an image.</td><td>Generate 3D PBR material maps from an image.</td></tr><tr><td><strong>Patina Material</strong></td><td>90</td><td>Generate a seamless tileable 3D PBR material from a text prompt.</td><td>Generate a seamless tileable 3D PBR material from a text prompt.</td></tr><tr><td><strong>Patina Material Extract</strong></td><td>120</td><td>Extract a seamless tileable 3D PBR material from a photo.</td><td>Extract a seamless tileable 3D PBR material from a photo.</td></tr><tr><td><strong>Qwen Image 2.0</strong></td><td>120</td><td>Next-gen unified editing model.</td><td>Next-gen unified editing model.</td></tr><tr><td><strong>Qwen Image 2.1</strong></td><td>60</td><td>Alibaba's unified text-to-image and editing model with strong typography.</td><td>Alibaba's unified text-to-image and editing model with strong typography.</td></tr><tr><td><strong>Qwen Image 3.0 Pro</strong></td><td>30</td><td>Alibaba's flagship generate-and-edit model: photoreal detail, precise text, up to 3 references.</td><td>Alibaba's flagship generate-and-edit model: photoreal detail, precise text, up to 3 references.</td></tr><tr><td><strong>Qwen Image Layered</strong></td><td>55</td><td>Split an image into multiple layers.</td><td>Split an image into multiple layers.</td></tr><tr><td><strong>Recraft V3</strong></td><td>25</td><td>Clean, vector-style illustrations and graphic design.</td><td>Concept art and creative storytelling</td></tr><tr><td><strong>Recraft V4</strong></td><td>40</td><td>High-quality image generation with enhanced detail.</td><td>Premium illustrations, detailed graphics, professional design assets</td></tr><tr><td><strong>Recraft V4 Pro</strong></td><td>60</td><td>High-quality image generation with enhanced detail.</td><td>Top-tier design outputs, maximum fidelity illustrations</td></tr><tr><td><strong>Recraft V4 Styles</strong></td><td>15</td><td>Style-consistent generation from your Recraft styles.</td><td>Style-consistent generation from your Recraft styles.</td></tr><tr><td><strong>Recraft V4 Styles Pro</strong></td><td>20</td><td>High-fidelity style-consistent generation from your Recraft styles.</td><td>High-fidelity style-consistent generation from your Recraft styles.</td></tr><tr><td><strong>Recraft V4.1</strong></td><td>30</td><td>Tuned for brand systems and editorial work.</td><td>Tuned for brand systems and editorial work.</td></tr><tr><td><strong>Recraft V4.1 Flash</strong></td><td>5</td><td>Fast, low-cost Recraft V4.1 for quick design iterations.</td><td>Fast, low-cost Recraft V4.1 for quick design iterations.</td></tr><tr><td><strong>Recraft V4.1 Pro</strong></td><td>60</td><td>Made for hero imagery, campaign work, and print.</td><td>Made for hero imagery, campaign work, and print.</td></tr><tr><td><strong>Recraft V4.1 Utility</strong></td><td>30</td><td>Optimized for quick, cost-effective image generation with a design focus.</td><td>Optimized for quick, cost-effective image generation with a design focus.</td></tr><tr><td><strong>Recraft V4.1 Utility Pro</strong></td><td>30</td><td>Premium-quality raster generation viable across full creative pipelines.</td><td>Premium-quality raster generation viable across full creative pipelines.</td></tr><tr><td><strong>Reve 2.1</strong></td><td>30</td><td>Reve 2.1 API Model</td><td>Reve 2.1 API Model</td></tr><tr><td><strong>Riverflow 2.0 Fast</strong></td><td>150</td><td>High quality image model with optimized latency.</td><td>Fast, high-quality generation with low wait times</td></tr><tr><td><strong>Riverflow 2.0 Pro</strong></td><td>245</td><td>Premium image model with high accuracy.</td><td>Professional outputs with maximum accuracy</td></tr><tr><td><strong>Riverflow 2.5 Pro</strong></td><td>245</td><td>Model suitable for commercial visual work that needs stronger output quality.</td><td>Model suitable for commercial visual work that needs stronger output quality.</td></tr><tr><td><strong>SAM 3</strong></td><td>10</td><td>Segment an image by describing what to isolate as a short noun phrase. Returns each matching object as its own cutout or black-and-white mask.</td><td>Segment an image by describing what to isolate as a short noun phrase. Returns each matching object as its own cutout or black-and-white mask.</td></tr><tr><td><strong>Seedream 3.0</strong></td><td>20</td><td>Cinematic film-style visuals with legible text.</td><td>Stylized marketing content, cinematic visuals</td></tr><tr><td><strong>Seedream 4.0</strong></td><td>50</td><td>Strong spatial understanding, good for batch work.</td><td>4K photoreal marketing visuals, consistent product-hero images, educational diagrams with precise labels</td></tr><tr><td><strong>Seedream 4.5</strong></td><td>65</td><td>Sharp text rendering with 4K output.</td><td>High-quality photoreal images, multi-image composition, 4K asset creation</td></tr><tr><td><strong>Seedream 5 Pro</strong></td><td>130</td><td>Efficient content creation and production capabilities</td><td>Efficient content creation and production capabilities</td></tr><tr><td><strong>Seedream 5 Pro Layerize</strong></td><td>100</td><td>Splits an image into independent, editable transparent-PNG layers.</td><td>Splits an image into independent, editable transparent-PNG layers.</td></tr><tr><td><strong>Seedream 5.0 Lite</strong></td><td>65</td><td>Next-gen image synthesis with enhanced prompt understanding.</td><td>Next-gen image synthesis with enhanced prompt understanding.</td></tr><tr><td><strong>Uni-1</strong></td><td>30</td><td>Edit an image using Luma's multimodal reasoning model.</td><td>Edit an image using Luma's multimodal reasoning model.</td></tr><tr><td><strong>Uni-1 Max</strong></td><td>30</td><td>Higher-quality unified image model by Luma.</td><td>Higher-quality unified image model by Luma.</td></tr><tr><td><strong>Wan 2.2</strong></td><td>25</td><td>High realism with fine detail control.</td><td>Cost-effective realistic image generation</td></tr><tr><td><strong>Z-Image Turbo</strong></td><td>20</td><td>Fast image-editing model with strength control.</td><td>Rapid iteration, budget-friendly generation, quick prototyping</td></tr></tbody></table>

## Editing Models

These models edit or transform existing images.

<table><thead><tr><th width="220">Model</th><th width="120">Est. time (s)</th><th>Description</th><th>Best for</th></tr></thead><tbody><tr><td><strong>Bria Guided Remove Background</strong></td><td>10</td><td>Cut out exactly what you name, keeping the original pixels.</td><td>Cut out exactly what you name, keeping the original pixels.</td></tr><tr><td><strong>Enhancor V1</strong></td><td>32</td><td>Portrait skin smoothing and refinement.</td><td>Subtle portrait retouching, skin smoothing</td></tr><tr><td><strong>Enhancor V3</strong></td><td>185</td><td>Professional-grade face enhancement.</td><td>Professional portrait retouching and enhancement</td></tr><tr><td><strong>Enhancor V4</strong></td><td>70</td><td>Professional-grade face enhancement.</td><td>Advanced face enhancement with latest capabilities</td></tr><tr><td><strong>FLUX 2 Pro Outpaint</strong></td><td>30</td><td>Extend an image past its edges, matching the existing scene.</td><td>Extend an image past its edges, matching the existing scene.</td></tr><tr><td><strong>Flux Canny</strong></td><td>30</td><td>Extract precise edges to maintain composition.</td><td>Precise composition control via edge detection for detailed scenes</td></tr><tr><td><strong>Flux Depth</strong></td><td>65</td><td>Create depth maps for 3D effects and spatial transforms.</td><td>Maintaining 3D spatial consistency and accurate perspectives</td></tr><tr><td><strong>Flux Kontext Max</strong></td><td>30</td><td>Complex scenes with strong spatial awareness.</td><td>Out-painting, style swaps, text edits, maintaining character continuity</td></tr><tr><td><strong>GPT Image 1.5 Inpainting</strong></td><td>60</td><td>OpenAI's powerful image inpainting model.</td><td>Precise inpainting with strong prompt comprehension</td></tr><tr><td><strong>Ideogram 4.5 Precise Edit</strong></td><td>30</td><td>Edit images turn after turn without drift. Untouched pixels stay exact.</td><td>Edit images turn after turn without drift. Untouched pixels stay exact.</td></tr><tr><td><strong>Ideogram Character</strong></td><td>30</td><td>Transfer faces between images.</td><td>Maintaining character identity across generations</td></tr><tr><td><strong>Nano Banana Pro Inpainting</strong></td><td>65</td><td>Highest quality from Google, up to 4K.</td><td>Premium inpainting with Google's best image model</td></tr><tr><td><strong>Qwen Image Edit</strong></td><td>25</td><td>Best for adding or changing text in images.</td><td>Text editing within images, precise inpainting</td></tr><tr><td><strong>Qwen Image Edit 2511 Angles</strong></td><td>24</td><td>Full 360 camera angle control with horizontal and vertical rotation.</td><td>Full 360 camera angle control with horizontal and vertical rotation.</td></tr><tr><td><strong>Qwen Image Edit Plus</strong></td><td>120</td><td>Complex layouts with multiple text elements.</td><td>Complex text edits, advanced image composition</td></tr><tr><td><strong>Remove background</strong></td><td>15</td><td>Isolate clean subjects from images.</td><td>Subject isolation, transparent backgrounds, compositing</td></tr><tr><td><strong>Riverflow 2.0 Pro Inpainting</strong></td><td>210</td><td>Premium image model with high accuracy.</td><td>Precise inpainting with high-accuracy results</td></tr></tbody></table>

## Vector & SVG Models

These models generate or convert vector graphics.

<table><thead><tr><th width="220">Model</th><th width="120">Est. time (s)</th><th>Description</th><th>Best for</th></tr></thead><tbody><tr><td><strong>Arrow 1.0</strong></td><td>500</td><td>Convert raster images to production-ready SVG vectors.</td><td>Raster-to-vector conversion, production-ready SVG output</td></tr><tr><td><strong>Arrow 1.0 References</strong></td><td>500</td><td>Text-to-SVG generation guided by up to 4 reference images.</td><td>Guided SVG creation with style and composition references</td></tr><tr><td><strong>Arrow 1.1</strong></td><td>500</td><td>Most efficient raster-to-vector model.</td><td>Faster raster-to-vector conversion with improved structure and quality.</td></tr><tr><td><strong>Arrow 1.1 Max</strong></td><td>500</td><td>Generate high quality SVG vectors using an image input.</td><td>Premium vector generation for detailed and complex illustrations.</td></tr><tr><td><strong>Arrow 1.1 Max References</strong></td><td>500</td><td>Highest quality text-to-SVG generation with reference image guidance.</td><td>Maximum quality reference-guided SVG generation.</td></tr><tr><td><strong>Arrow 1.1 References</strong></td><td>500</td><td>Updated and more efficient text-to-SVG model guided by image references.</td><td>Reference-guided SVG generation with improved speed and consistency.</td></tr><tr><td><strong>Arrow 2 Telos</strong></td><td>90</td><td>Quiver's most capable SVG model: Arrow with a frontier model's refinement, for complex styles, compositions and briefs.</td><td>Quiver's most capable SVG model: Arrow with a frontier model's refinement, for complex styles, compositions and briefs.</td></tr><tr><td><strong>Arrow 2 Telos References</strong></td><td>90</td><td>Quiver's most capable text-to-SVG generation, refined by a frontier model and guided by up to fourteen reference images.</td><td>Quiver's most capable text-to-SVG generation, refined by a frontier model and guided by up to fourteen reference images.</td></tr><tr><td><strong>Arrow 2.0</strong></td><td>90</td><td>Quiver's flagship SVG model: cleaner layering, palettes and outlines, balancing quality and speed.</td><td>Quiver's flagship SVG model: cleaner layering, palettes and outlines, balancing quality and speed.</td></tr><tr><td><strong>Arrow 2.0 References</strong></td><td>90</td><td>Quiver's flagship text-to-SVG generation, guided by up to fourteen reference images.</td><td>Quiver's flagship text-to-SVG generation, guided by up to fourteen reference images.</td></tr><tr><td><strong>Recraft V4 Pro Vector</strong></td><td>55</td><td>High-quality vector illustration generation.</td><td>Premium vector illustrations, maximum fidelity</td></tr><tr><td><strong>Recraft V4 Styles Pro Vector</strong></td><td>20</td><td>High-fidelity style-consistent vector (SVG) generation from your Recraft vector styles.</td><td>High-fidelity style-consistent vector (SVG) generation from your Recraft vector styles.</td></tr><tr><td><strong>Recraft V4 Styles Vector</strong></td><td>15</td><td>Style-consistent vector (SVG) generation from your Recraft vector styles.</td><td>Style-consistent vector (SVG) generation from your Recraft vector styles.</td></tr><tr><td><strong>Recraft V4 Vector</strong></td><td>40</td><td>Vector illustrations with enhanced detail.</td><td>Detailed vector illustrations and graphic design</td></tr><tr><td><strong>Recraft V4.1 Pro Vector</strong></td><td>55</td><td>High-quality vector illustration generation.</td><td>High-quality vector illustration generation.</td></tr><tr><td><strong>Recraft V4.1 Vector</strong></td><td>40</td><td>Built for logos, icons, and illustration systems</td><td>Built for logos, icons, and illustration systems</td></tr></tbody></table>

## Upscaling & Enhancement

These models upscale resolution or enhance image quality.

<table><thead><tr><th width="220">Model</th><th width="120">Est. time (s)</th><th>Description</th><th>Best for</th></tr></thead><tbody><tr><td><strong>Magnific Creative Upscaler</strong></td><td>95</td><td>Artistic upscaling with creative options.</td><td>Creative upscaling with stylistic enhancement</td></tr><tr><td><strong>Magnific Precision Upscaler</strong></td><td>65</td><td>Detail-preserving 2× upscaling.</td><td>Sharp, faithful upscaling that preserves original detail</td></tr><tr><td><strong>Magnific Precision Upscaler V2</strong></td><td>60</td><td>Sharp edges and texture recovery.</td><td>Maximum detail recovery, edge-preserving upscaling</td></tr><tr><td><strong>Topaz Generative Upscaler</strong></td><td>60</td><td>Generative upscaling, reconstructs missing detail, faces, and textures.</td><td>Generative upscaling, reconstructs missing detail, faces, and textures.</td></tr><tr><td><strong>Topaz Upscaler</strong></td><td>25</td><td>Boost resolution and detail, industry-standard.</td><td>Boost resolution and detail, industry-standard.</td></tr></tbody></table>


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.flora.ai/models/image-models.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.