AI models

CreativeArc uses a curated set of AI models, each chosen for a specific task. You can switch between models where alternatives are available using the Model dropdown in the relevant tab.

Image generation

Nano Banana Pro

Google's high-quality photorealistic image generation and transformation across all design tabs. Runs at 1K, 2K, and 4K.

Nano Banana 2

Google's fast image generation-same capabilities as Nano Banana Pro at lower cost, at 1K, 2K, and 4K.

GPT-Image 2

OpenAI-powered image generation and editing-an alternative model for the Image tab.

Flux-2 Pro

Black Forest Labs' image generation model-an alternative style and quality profile for the Image tab.

Flux-2 Max

Black Forest Labs' highest-fidelity image generation model.

Dola Seedream 5.0 Pro

BytePlus Ark's image generation model, at 1.5K and 4K-also used for transparent-background product cutouts and for splitting a design into editable layers.

Recraft V4.1

Recraft's raster image generation in Vector mode-Standard and Pro (higher resolution and detail), plus Utility and Utility Pro for simpler, more predictable output.

Vector artwork

Recraft V4.1 Vector

True SVG generation from a prompt-Vector, Pro Vector, Utility Vector, and Utility Pro Vector variants, each matching its raster counterpart's style profile.

Recraft Vectorize

One-click conversion of an existing image into a true SVG.

Text & reasoning

GPT-5.6 Terra

AI assistant, prompt enhancement, image description, furniture detection, and design brief parsing.

Claude Sonnet 5

Anthropic's model behind the Interactive assistant-plans and applies edits to a deck. Priced per turn from the tokens that turn actually used.

Live Search

Google Search grounding-pulls real products and reference material into a generation. Billed per search query on top of the generation itself.

Video

Veo 3.1

Google's video generation model for animating renders with natural motion and lighting, at 720p, 1080p, and 4K.

Kling V3

Alternative video generation model-cinematic motion from still images or prompts, with first- and last-frame control.

Flux 3

Black Forest Labs' video generation model-text-to-video, image-to-video, and video-to-video, each at draft, HD, and FHD quality.

Dreamina Seedance 2.5

ByteDance's video generation model, at 480p, 720p, and 1080p from either a prompt or an existing clip.

Gemini Omni Flash

Google's video-to-video editing model-apply prompted edits to existing footage.

Upscaling

Clarity Pro Upscaler

AI super-resolution at 2×, 4×, 8×, and 16×-sharpens detail and recovers texture without hallucination. Includes a creativity slider to balance faithfulness vs. enhancement.

Topaz Image Upscale

AI super-resolution at 2×, 4×, and 6× via Topaz Labs. Five processing modes-Standard V2, Low Resolution V2, CGI, High Fidelity V2, Text Refine-plus optional face enhancement with creativity and strength controls.

Magnific Creative

Prompt-guided upscaling at 2×, 4×, and 8–16×-HDR, resemblance, and fractality sliders with selectable engines and flavors, for reimagining detail rather than preserving it.

Magnific Precision

Faithful upscaling at 2×, 4×, and 8–16× using a curated set of Magnific v1/v2 flavors, with sharpen, grain, and detail sliders.

P-Image Upscale

AI super-resolution at 2×, 4×, 6×, and 8×-a lower-cost alternative for straightforward upscales. Priced by the megapixels of the finished image.

Flux Deblur

Sharpens and removes blur from an image without changing its resolution.

Flux 3 Upscale

Upscales generated video at Precise or Creative quality levels.

Topaz Video Upscale

Video super-resolution to 720p, 1080p, or 4K at 60fps.

Background & object removal

BiRefNet

Precise subject isolation-removes backgrounds cleanly for compositing or export.

Recraft Remove BG

Recraft's background remover-the one-click Remove BG action in Vector mode.

Flux Erase

Removes a specific object from an image while preserving the rest of the scene.

3D generation

Hunyuan 3D 3.1

Fast, high-quality 3D model generation from images or text descriptions.

Tripo v3.1

High-fidelity 3D generation from an image, a text prompt, or multiple views, with texture tiers up to 8K Ultra plus retopology, quad mesh, segmentation, and part completion add-ons.

Meshy 7

3D model generation from images or prompts, untextured or with 2K, 4K, or 8K texture, with an optional Ultra quality tier.

Meshy Low Poly

Low-polygon 3D generation-lighter meshes for real-time scenes and quick blocking.

Meshy Smart Topology

Generation with cleaned-up, animation-friendly edge flow.

Meshy Remesh

Rebuilds the geometry of an existing model at a different polygon budget or topology.

Meshy Retexture

Applies new AI-generated 2K, 4K, or 8K textures to a model you already have.

World generation

Marble-1.1

World Labs' standard world generation model-converts images, panoramas, video, or text into fully explorable immersive 3D worlds.

Marble-1.1-plus

World Labs' highest-quality world generation model-slower but produces richer geometry, lighting, and detail.

Panorama

Generates the 360° equirectangular panorama every world is built from, unless your source is already one. Included in the cost of the world.

Data & privacy: When you run a generation, your prompt and any uploaded images are transmitted to the relevant AI provider to process your request. This is covered in full in the Privacy Policy and in the Legal-Data, Security & IP section of these docs.

Models are updated as better options become available. Your generations are always processed by whichever model version is current-you don't need to do anything when models are upgraded.