Feb 18, 10:48 PM

StepFun AI Launches 30 Billion Parameter Text-to-Video Model Step-Video-T2V at 540p Resolution, Generating Up to 204 Frames

StepFun AI has launched a new open-source text-to-video model named Step-Video-T2V, featuring 30 billion parameters and the capability to generate up to 204 frames per video. This model is released under the MIT license, promoting collaboration within the AI community. It is designed to enhance video generation with high compression, allowing for large-scale creation without significant quality loss. Additionally, the model supports 540p resolution and is trained on thousands of H800s, positioning it competitively against existing models like Meta's MovieGen. The release is part of a broader trend in AI, as various companies unveil advancements in video generation technology, including Topaz Labs' Project Starlight, which focuses on video enhancement using diffusion models.

#StepFun AI #MIT #Meta #MovieGen #AI #Topaz Labs #Project Starlight

Written with ChatGPT (GPT-4o mini).

Sources

Additional media

Image #1 for story stepfun-ai-launches-30-billion-parameter-text-to-video-model-step-video-t2v-540p-0d6ecc57

Image #2 for story stepfun-ai-launches-30-billion-parameter-text-to-video-model-step-video-t2v-540p-0d6ecc57

Image #3 for story stepfun-ai-launches-30-billion-parameter-text-to-video-model-step-video-t2v-540p-0d6ecc57

Image #4 for story stepfun-ai-launches-30-billion-parameter-text-to-video-model-step-video-t2v-540p-0d6ecc57

Image #5 for story stepfun-ai-launches-30-billion-parameter-text-to-video-model-step-video-t2v-540p-0d6ecc57