A very fast and cheap PrunaAI optimized version of Wan 2.2 A14B text-to-video
Frequently asked questions
What is wan-2.2-t2v-fast and how does it work?
wan-2.2-t2v-fast is a speed-optimized text-to-video AI model built on the Wan 2.2 A14B architecture and compressed by PrunaAI for faster, cheaper inference. It converts natural language prompts in English or Chinese directly into short video clips at 480p or 720p resolution, with automatic FPS interpolation to 30 FPS via ffmpeg.
How fast is wan-2.2-t2v-fast compared to standard Wan 2.2?
wan-2.2-t2v-fast is significantly faster and cheaper than the standard Wan 2.2 baseline. PrunaAI's state-of-the-art model compression algorithms reduce compute cost while preserving acceptable visual quality, making it the preferred choice for rapid video prototyping and high-volume generation workflows.
What resolutions and aspect ratios does wan-2.2-t2v-fast support?
The model supports 480p and 720p resolutions with two aspect ratios: 16:9 landscape (832×480px) for widescreen content and 9:16 portrait (480×832px) for mobile-first or vertical video. This flexibility makes it suitable for social media, web, and app content creation across different platforms.
Does wan-2.2-t2v-fast support LoRA adapters for custom styles?
Yes. wan-2.2-t2v-fast supports loading up to two independent transformer LoRA adapters via URL. This allows creators and developers to apply custom stylistic fine-tuning on top of the base model, enabling branded or artistically consistent video output without retraining the full model.
How many frames can wan-2.2-t2v-fast generate per video?
wan-2.2-t2v-fast supports a variable frame count of up to 81 frames per generation. Generating the maximum 81 frames delivers the highest visual quality output. The model also automatically interpolates the generated frames to 30 FPS using ffmpeg, ensuring smooth and professional-looking playback.
Can wan-2.2-t2v-fast generate videos from Chinese prompts?
Yes. The model includes a built-in optional Chinese prompt translation feature that translates prompts before generation, leveraging Wan 2.2's native Chinese language strengths. This makes it highly effective for Chinese-language creators and teams who want to write prompts in their native language.
Who created wan-2.2-t2v-fast and where is it hosted?
wan-2.2-t2v-fast was created by wan-video and optimized by PrunaAI using advanced model compression techniques. It is hosted on Replicate and available on CRAISEE. As of the latest data, the model has accumulated over 361,712 runs, reflecting strong adoption among developers and content creators.
What are the best use cases for wan-2.2-t2v-fast?
wan-2.2-t2v-fast is best suited for rapid video prototyping, iterating across multiple video concepts quickly, high-volume low-cost video generation, and workflows where speed matters more than maximum quality. It is ideal for developers, indie creators, and teams on CRAISEE who need fast turnaround before committing to higher-fidelity models.

