Wan2.2-TI2V-5B

Wan2.2-TI2V-5B (Wan-AI)

Text-and-image to video generation with a smaller, more practical footprint.

• Dual-input pipeline. Combines a text prompt with a reference image to generate short videos, helping anchor style, characters, and composition from the start.

• Mid-size diffusion model. ~5B parameters, offering a balance between visual quality and compute compared to the larger 14B Wan models.

• More control, less drift. Using an input image reduces subject and style drift across frames, especially for character-focused scenes.

• Diffusers-ready. Packaged for the Diffusers ecosystem, making integration and experimentation straightforward.

• Research-oriented release. Intended for evaluation and prototyping rather than real-time or consumer-grade deployment.

Why pick it for Norman AI?

Wan2.2-TI2V-5B is a good middle ground when you want guided video generation without extreme compute costs. It’s useful for testing text-plus-image conditioning, character consistency, and motion quality in practical video workflows.

response = await norman.invoke(
    {
        "model_name": "wan2.2-ti2v-5b",
        "inputs": [
            {
                "display_title": "Prompt",
                "data": "A cat playing with a ball on mars"
            }
        ]
    }
)