vllm-omni-video-gen

Generate video from text prompts or reference images using Wan2.2 by running T2V, I2V, or TI2V workflows through vLLM-Omni. You provide a text prompt and an optional reference image to produce high-quality video content.

Yes, Wan2.2 diffusion models support text-to-video (T2V), image-to-video (I2V), and text+image-to-video (TI2V) generation. vLLM-Omni applies these workflows across multiple Wan2.2 variants to create videos from your inputs.

Available Wan2.2 model variants include Wan2.2-T2V-A14B, Wan2.2-I2V-A14B, Wan2.2-TI2V-5B, and NextStep-1.1. You can choose from these models to match your specific hardware constraints and VRAM limits.

Control video output quality and performance during diffusion by configuring parameters such as num_inference_steps, guidance_scale, seed, fps, and num_frames. Specifying these settings adjusts the final rendering quality and generation speed.

VRAM requirements for running Wan2.2 video generation models vary depending on your selected variant, with specific VRAM guidance provided for Wan2.2-T2V-A14B, Wan2.2-I2V-A14B, and Wan2.2-TI2V-5B to match your hardware constraints.

Set up vLLM-Omni for Wan2.2 video generation by consulting the detailed model specifications and setup instructions in the references. You configure model choice, inference steps, and frame rate before executing the generation workflow.