from vllm-omni-skills50
Generate videos (text→video, image→video, text+image→video) using vLLM-Omni and Wan2.2-style diffusion models, with guidance on parameters and performance trade
This skill documents how to use vLLM-Omni with Wan2.2 and related models to generate videos in three modes: text-to-video (T2V), image-to-video (I2V), and text+image-to-video (TI2V). It provides quick-start code samples for offline and API usage, model IDs, recommended VRAM, common generation parameters (steps, guidance scale, frames, fps), and troubleshooting tips for memory and performance.
Use it when you need programmatic video generation from prompts or reference images, when experimenting with diffusion transformer models for motion, or when building pipelines that convert text/image inputs into short video outputs. Suitable for researchers and engineers with GPU resources (24–48GB VRAM for larger models).
references/wan-models.md (has_references=true)Inferred: code and ML-focused agents (Copilot, Codex, Claude-Code) and orchestration tooling that can run vLLM or serve models behind APIs.
Documentation-only skill for generating videos via vLLM-Omni using Wan2.2 diffusion models. Covers T2V, I2V, and TI2V modes with clear code examples, parameter tables, and performance tips. No scripts bundled — purely reference/instructional. Clean SKILL.md with good structure and specific triggers, but limited to users with high-VRAM GPUs (24-48GB minimum).
No security concerns — purely documentation with safe code examples. Niche audience due to heavy hardware requirements. Well-written but would benefit from bundled helper scripts for common tasks like batch generation or parameter sweeps.
