SkyReels

Wan2.2 is a major upgrade to the Wan suite of open video foundation models, introducing a Mixture‑of‑Experts (MoE) architecture that splits the diffusion denoising process across high‑noise and low‑noise expert paths to dramatically increase model capacity without raising inference cost. It harnesses meticulously labeled aesthetic data, covering lighting, composition, contrast, and color tone, to enable precise, controllable cinematic‑style video generation. Trained on over 65 % more images and 83 % more videos than its predecessor, Wan2.2 delivers top performance in motion, semantic, and aesthetic generalization. The release includes a compact, high‑compression TI2V‑5B model built on an advanced VAE with a 16×16×4 compression ratio, capable of text‑to‑video and image‑to‑video synthesis at 720p/24 fps on consumer GPUs such as the RTX 4090. Prebuilt checkpoints for T2V‑A14B, I2V‑A14B, and TI2V‑5B stack enable seamless integration.

FramePack AI revolutionizes video creation by enabling the generation of long, high-quality videos on consumer GPUs with just 6 GB of VRAM, using smart frame compression and bi-directional sampling to maintain constant computational load regardless of video length while avoiding drift and preserving visual fidelity. Key innovations include fixed context length to compress frames by importance, progressive frame compression for optimal memory use, and anti-drifting sampling to prevent error accumulation. Fully compatible with existing pretrained video diffusion models, FramePack accelerates training with large batch support and integrates seamlessly via fine-tuning under an Apache 2.0 open source license. Its user-friendly workflow lets creators upload an image or initial frame, set preferences for length, frame rate, and style, generate frames progressively, and preview or download final animations in real time.