One of the great things about Layer is the fact that we’re model-agnostic, meaning we support many AI models with more coming in all the time. With AI evolving on almost a weekly basis, there’s no need to get locked into any one model - simply use the best model for your usecase.
This approach ensures that game studios can access the best-in-class tools without being locked into a single provider. Below, we break down the different AI models Layer supports across three categories: 2D, 3D, and Video.
2D Image Generation Models
For creating high-quality 2D game assets like characters, concept art, backgrounds, and more. These can be trained in your own custom style and generated in large batches.
FLUX: Layer’s most popular AI model for stylized and high-quality 2D asset generation.
FLUX.1 Fast
FLUX.1 Dev
Supports custom styles
Stable Diffusion: An open-source model ideal for flexible and customizable image generation. The majority of base Layer styles run on Stability AI.
Stable Diffusion 3
Stable Diffusion XL
Supports custom styles.
BRIA: Designed for professional-grade visual content with ethical copyright compliant sourcing. Available for Enterprise users.
Supports custom styles.
Google Imagen: A high-fidelity image generation model from Google’s AI research.
Google Gemini: A versatile multimodal model capable of high-quality image and video generation, with strong performance in complex scene creation and editing.
Qwen: Alibaba’s latest model for image generation and editing, optimized for flexibility, refinement, and high-accuracy asset creation.
DALL·E 3: OpenAI’s latest model for detailed and context-aware image generation.
3D Asset Generation Models
For detailed 3D assets you can bring into major game engines like Unity or Unreal.
Tripo AI: Specialized in fast 3D shape and object generation for gaming applications.
Trellis 3D: Microsoft-designed model optimized for procedural 3D content creation.
Rodin: Meta’s AI model focused on generating high-quality 3D assets from both text and image prompts.
Hunyuan3D: Developed by Tencent, Hunyuan3D can generate high-quality 3D models from image prompts.
Video Generation Models
For creating video assets like trailers, game items, UA/marketing ads, or animated backgrounds - Layer has wide range of video models available.
Minimax: Video model optimized for smooth AI-generated motion.
Minimax Live: Real-time variation of Minimax, enabling faster live animation generation.
Framepack*: Compact architecture—achieves high-quality output with fewer compute resources and longer videos. Quality is compromised.
Higgsfield AI: Best for short-form video content with a high degree of personalization.
Higgsfield Turbo: An upgraded version of Higgsfield AI, delivering faster results and sharper motion for short-form video.
Hunyuan*: Powerful AI model from Tencent for high-quality video synthesis.
Kling Pro 1.5: High-resolution AI model built for detailed AI video generation.
Kling Pro 1.6: The latest iteration of Kling’s AI model with enhanced video fidelity.
Kling 2.1 Master: Kling’s newest flagship model, designed for ultra-high fidelity video generation with advanced motion handling and cinematic quality.
Veo 2: Google DeepMind’s advanced video generation model.
Veo 3: Google DeepMind’s newest video model, offering greater realism, longer durations, and higher fidelity across cinematic and game-ready assets.
Wan 2.1*: New from Alibaba, Wan is efficient and consistent, with a cutting-edge diffusion architecture + features like 3D causal VAE encoding for sharper outputs.
Seedance 1.0: Bytedance’s video generation model, optimized for natural motion and strong character animation in short-to-mid length clips.
Magi Distilled: A distilled video generation model built for efficiency — delivering faster outputs while retaining strong visual quality.
AI Avatar: Specialized for character-driven video, enabling personalized avatars and digital humans for games, trailers, and live content.
*The models marked with (*) has seed editing supported for keeping videos cohesive while making subtle changes on the prompt.
Layer’s model-agnostic approach means that gamedevs can leverage the best AI tools available without worrying about constant integration and optimization headaches. Whether you’re building things like detailed 2D environments, generating 3D models to bring into Unreal or Unity, or crafting videos ads for UA campaigns - Layer always has your back.