✨ GPT Image 1.5 Edit is now live! ✨
← Back to Models
🎥 Video Generation

Wan v2.6 Reference-to-Video

Wan 2.6 reference-to-video model. Maintain subject consistency across scenes using 1-3 reference videos. Reference subjects as @Video1, @Video2, @Video3 in prompts. Works for people, animals, objects

Example Output

Prompt

"Dance battle between @Video1 and @Video2"

Generated Result

Generated

Try Wan v2.6 Reference-to-Video

Fill in the parameters below and click "Generate" to try this model

Reference videos for subject consistency (1-3 videos). Reference in prompt as @Video1, @Video2, @Video3

Prompt with @Video1, @Video2, @Video3 references (max 800 chars). For multi-shot: '[0-3s] Shot 1. [3-6s] Shot 2.'

Video aspect ratio

Video resolution (R2V supports 720p and 1080p only)

Video duration (R2V supports 5 or 10 seconds only)

Enable LLM prompt rewriting

Enable multi-shot segmentation for coherent narrative (only when prompt expansion enabled)

Content to avoid (max 500 chars)

Your inputs will be saved and ready after sign in

More Video Generation Models

Pika v2.2 PikaScenes

Combine multiple images into a single 5-second video with creative or precise blending.

LTX Video 2.0 Pro Image To Video

Transform images into cinematic 4K videos with audio.

SkyReels V1 Image to Video

Create cinematic videos from your images with film-quality motion and detail.

VEED Fabric 1.0 Text

VEED Fabric 1.0 Text

Turn text and images into talking avatar videos with auto lip-sync and natural voice generation.

CogVideoX-5B Image to Video

CogVideoX-5B Image to Video

Animate images with natural motion using text prompts to guide the action.

Google Veo 3.1 First-Last-Frame

Create videos with smooth transitions between two keyframes.

Wan v2.6 Image-to-Video

Wan 2.6 image-to-video model. Animate images with text prompts, supports multi-shot generation and background audio. Image size: 360-2000px, max 100MB

PixVerse v4.5 Text-to-Video

Create video clips from text descriptions up to 8s long in 1080p

Wan 2.2 Animate Replace

Replace characters in videos while keeping original lighting and scene intact.

About Wan v2.6 Reference-to-Video

Wan v2.6 Reference-to-Video is an advanced AI video generation model designed to maintain subject consistency across multiple scenes by leveraging 1 to 3 reference videos. With this cutting-edge technology, users can create visually coherent videos featuring people, animals, or objects, ensuring that the identity and appearance of the main subjects remain intact throughout the generated footage. The model stands out by allowing users to specify references as @Video1, @Video2, and @Video3 directly in their prompts, making it easy to direct the narrative and visual focus of each shot. Wan v2.6 is built for flexibility and ease of use. Users can upload up to three reference videos, which the AI uses to understand and replicate specific subjects in new, AI-generated video content. The prompt system allows for detailed scene descriptions, including multi-shot segmentation, where users can divide the video timeline into distinct segments (e.g., '[0-3s] Shot 1. [3-6s] Shot 2.'). This enables the creation of complex, multi-part stories with smooth transitions and logical narrative flow, all while preserving subject consistency. The model supports popular aspect ratios such as 16:9 (landscape), 9:16 (portrait), 1:1 (square), 4:3, and 3:4, providing flexibility for different platforms and creative needs. Resolution options include 720p HD and 1080p Full HD, ensuring crisp and professional video output. Users can choose between 5-second and 10-second video durations, catering to both short-form and slightly longer content requirements. Wan v2.6 incorporates powerful features like LLM-based prompt expansion, which refines and enhances user prompts for even better results. Multi-shot segmentation further elevates narrative coherence, making it a top choice for content creators who want to tell engaging stories or demonstrate actions across multiple scenes. The negative prompt option lets users specify unwanted elements, helping to fine-tune outputs for quality and relevance. Additional controls such as random seed settings for reproducibility and an integrated safety checker make the model both reliable and safe for a wide range of applications. Ideal use cases include social media content creation, marketing videos, storytelling, product showcases, and educational materials. Whether you're animating a dance battle between two people, showcasing products with consistent branding, or telling a story with recurring characters or objects, Wan v2.6 Reference-to-Video streamlines the process and elevates the creative potential of AI-driven video generation. This model is particularly well-suited for professional designers, marketers, video editors, educators, and anyone looking to produce high-quality, consistent video content quickly and effortlessly. With its intuitive interface, robust subject referencing, and advanced AI capabilities, Wan v2.6 sets a new standard for reference-based video generation, empowering users to bring their ideas to life with precision and creativity.

✨ Key Features

Maintains subject consistency across scenes using up to three reference videos for seamless storytelling.

Supports detailed prompts with direct references (@Video1, @Video2, @Video3) and multi-shot segmentation for complex narratives.

Flexible aspect ratio and resolution options, including 16:9, 9:16, 1:1, 4:3, 3:4, and up to 1080p Full HD output.

Intelligent LLM-based prompt expansion refines user input for improved video quality and coherence.

Negative prompt functionality allows users to exclude unwanted elements from generated videos.

Quick video generation with selectable durations of 5 or 10 seconds, ideal for social media and promotional content.

Integrated safety checker and reproducibility controls for safe, reliable content creation.

💡 Use Cases

Creating social media videos featuring recurring characters, animals, or objects with consistent appearance.

Producing marketing ads or product showcases where brand identity and subject consistency are crucial.

Developing educational or training videos featuring the same instructor or demonstrator across scenes.

Storytelling or short-form video projects that require seamless transitions between multiple shots.

Animating dance battles, sports actions, or creative performances using reference footage.

Generating personalized greeting videos or interactive content with familiar faces or mascots.

Enhancing video editing workflows by automating the generation of consistent visual elements.

🎯

Best For

Professional designers, marketers, content creators, educators, and video editors seeking consistent, high-quality AI-generated videos.

👍 Pros

  • Ensures subject consistency across all shots for visually coherent videos.
  • Supports complex, multi-scene narratives with advanced prompt and segmentation controls.
  • Offers flexible output settings for various platforms and creative requirements.
  • Streamlines creative workflows, reducing manual editing and production time.
  • Easy-to-use interface suitable for both beginners and professionals.
  • Incorporates safety and reproducibility features for secure and reliable use.

⚠️ Considerations

  • Supports only 5 or 10-second video durations, limiting longer content creation.
  • Requires clear reference videos for optimal subject consistency.
  • Currently limited to 720p and 1080p resolution options.
  • Multi-shot segmentation is only available when prompt expansion is enabled.

📚 How to Use Wan v2.6 Reference-to-Video

1

Prepare 1 to 3 reference videos of the subjects you want to feature in your generated video.

2

Upload your reference videos and assign them as @Video1, @Video2, and @Video3 in your prompt.

3

Write a detailed prompt, including scene descriptions and, if desired, multi-shot segmentation (e.g., '[0-3s] Shot 1. [3-6s] Shot 2.').

4

Select your preferred aspect ratio, video resolution, and duration from the available options.

5

Optionally, enable prompt expansion and multi-shots for enhanced narrative control, and use the negative prompt field to exclude unwanted content.

6

Submit your input and wait for the AI to generate your video, then download and review the final result.

Frequently Asked Questions

🏷️ Related Keywords

AI video generation reference-to-video subject consistency video AI tools creative storytelling multi-shot video content creation video editing AI automated video production LLM prompt expansion