Introducing WaveSpeedAI Uno on WaveSpeedAI

Introducing UNO: ByteDance’s Revolutionary Subject-Driven Image Generation Model Now on WaveSpeedAI

The challenge of maintaining character and object consistency across AI-generated images has long been a frustrating limitation for creators, marketers, and developers alike. Today, we’re excited to announce that UNO—ByteDance Research’s groundbreaking Universal In-Context Diffusion Transformer—is now available on WaveSpeedAI, bringing state-of-the-art subject-driven image generation to your fingertips with instant API access.

Whether you’re building a comic series, generating e-commerce product shots, or creating consistent brand mascots, UNO solves the “prosopagnosia” problem that has plagued AI image generation since its inception. Your subjects will finally look like themselves across every generated image.

What is UNO?

UNO (Universal In-Context Diffusion Transformer) is a subject-driven image generation framework developed by ByteDance’s Creative Intelligence team. Accepted to ICCV 2025, UNO represents a fundamental advancement in how AI handles visual identity—enabling the creation of new images where subjects from your reference photos reappear with high identity consistency and strong style control.

Built on the proven FLUX.1 architecture, UNO introduces two key innovations that set it apart:

Progressive Cross-Modal Alignment: A sophisticated two-stage training approach that first teaches the model single-subject consistency, then scales to complex multi-subject scenarios
Universal Rotary Position Embedding (UnoPE): A novel mechanism that helps the model’s attention distinguish between different visual sources, dramatically reducing the attribute confusion that plagues competing solutions

The result? A model that achieves state-of-the-art scores on DreamBench for subject similarity metrics while maintaining highly competitive text fidelity.

Key Features

Unmatched Subject Consistency

Keep the same person, character, or product instantly recognizable across unlimited new scenes, poses, and contexts
Maintain precise identity features including facial characteristics, clothing details, and distinctive accessories
Works with people, products, mascots, characters, and virtually any visual subject

Single to Multi-Subject Generation

Start with one subject or combine up to 5 reference images in a single generation
Create coherent group scenes with multiple subjects interacting naturally
Each subject maintains its unique identity without attribute bleeding or confusion

Flexible Creative Control

Guide compositions with natural language prompts describing desired scenes and styles
Support for multiple aspect ratios: square, portrait (4:3, 16:9), and landscape formats
Fine-tune outputs with adjustable guidance scale and inference steps
Reproducible results with optional seed control

Production-Ready Performance

Generates high-quality images at just $0.05 per image
No cold starts—instant inference on WaveSpeedAI’s optimized infrastructure
Simple REST API integration for seamless workflow automation

Real-World Use Cases

E-Commerce Product Photography

Transform a single product photo into dozens of lifestyle shots, seasonal campaigns, and contextual scenes. Generate your product in a minimalist studio setting, then in a cozy home environment, then on a sun-drenched beach—all while maintaining perfect product fidelity. No expensive photoshoots required.

Character-Consistent Content Creation

Comic artists, storyboard designers, and game developers can finally create extended visual narratives where protagonists look the same from panel to panel. Generate your hero in action poses, emotional close-ups, and wide establishing shots without manual character redesign.

Brand Asset Generation

Marketing teams can produce consistent brand mascot appearances across social media posts, advertising campaigns, and promotional materials. Your brand character will maintain its identity whether it’s celebrating a holiday, launching a product, or engaging with customers.

Virtual Try-On and Fashion

Show clothing and accessories on consistent model representations. Generate the same virtual model wearing different outfits or in various settings, creating cohesive lookbooks and product catalogs.

Rapid Concept Exploration

Concept artists and designers can quickly iterate on visual ideas while maintaining specific character or object designs. Explore dozens of compositional variations without losing the core identity elements that make your concepts unique.

Getting Started on WaveSpeedAI

Integrating UNO into your workflow is straightforward with WaveSpeedAI’s REST API:

Upload Reference Images: Provide 1-5 images of your subject(s). Use multiple angles or expressions for enhanced consistency.
Craft Your Prompt: Describe the scene you want to generate. Be specific about setting, action, and style—UNO will combine your text direction with reference identity.
Configure Parameters: Choose your aspect ratio (square_hd, portrait_16_9, landscape_4_3, etc.), set your desired number of outputs, and optionally specify a seed for reproducibility.
Generate: Call the API and receive your subject-consistent images in seconds, ready for immediate use.

API Highlights

Endpoint: https://wavespeed.ai/models/wavespeed-ai/uno
Cost: $0.05 per generated image
Inputs: 1-5 reference images + text prompt
Outputs: JPEG or PNG in multiple aspect ratios

WaveSpeedAI’s infrastructure eliminates cold starts entirely, meaning your first request runs just as fast as your thousandth. Whether you’re generating a single hero image or batch-processing thousands of product variants, you’ll experience consistent, production-grade performance.

Why Choose WaveSpeedAI for UNO?

Running UNO locally requires substantial GPU resources—approximately 16GB VRAM even in optimized fp8 mode. WaveSpeedAI removes this barrier entirely:

Zero Infrastructure Management: No GPU provisioning, no model weight downloads, no dependency conflicts
Instant Availability: Skip the cold start delays that plague other inference platforms
Predictable Pricing: Simple per-image billing at $0.05 with no hidden costs
Production Reliability: Enterprise-grade uptime for mission-critical applications
Easy Integration: Clean REST API with comprehensive documentation

Transform Your Visual Content Pipeline

UNO represents a genuine leap forward in AI image generation. By solving the subject consistency challenge, it unlocks creative possibilities that were previously impractical or impossible—from character-driven storytelling to scalable product visualization.

The combination of ByteDance’s cutting-edge research and WaveSpeedAI’s optimized inference infrastructure means you can start leveraging these capabilities immediately, without the complexity of self-hosting or the unpredictability of cold-start delays.

Ready to experience subject-consistent image generation? Visit UNO on WaveSpeedAI to explore the API documentation, try sample generations, and integrate UNO into your creative pipeline today.