🧵Eternal AI Integrates Wan 2.2 with Filecoin In Short: Eternal AI uses Wan 2.2 for uncensored video generation and stores its models on Filecoin via Lighthouse for open access. The system pairs large models for inference, text-to-video, and image-to-video generation. 1/ Core Model @EternalAI_ has integrated Wan 2.2, an open-source video model developed by @Alibaba_Wan, as the foundation of its uncensored video system. Users can input a prompt or a photo to generate short, cinematic videos containing NSFW content. 2/ Model Architecture The Wan 2.2 release includes separate models for text-to-video and image-to-video generation. Eternal AI uses the Wan2.2-T2V-A14B model for prompts and the Wan2.2-I2V-A14B model for visuals. Each model is built with a 14B parameter architecture optimized for cinematic motion. 3/ Decentralized Access To support free, permanent access to these models, Eternal AI uploaded Wan 2.2 to decentralized storage on Filecoin via @LighthouseWeb3. The IPFS CID is publicly available, allowing anyone to directly retrieve the model. 4/ Other Models In addition to Wan 2.2, Eternal AI also uses the Hermes-4-70B model from @NousResearch, a large-scale language model for prompt generation and inference. This model prepares the instructions and interacts with the video backend to guide the output. 5/ No Local Setup Required The system requires no local setup. Users simply upload an image, enter a prompt, and receive a video response. The service operates in an open mode and relies on decentralized storage to maintain access, even if the website disappears. Transform simple prompts or photos into short, cinematic videos, even NSFW ones: