Genmo Unveils Mochi 1: A New Chapter in AI Video Generation
AI company Genmo has just released a groundbreaking open-source model, Mochi 1, aimed at transforming video generation through AI. Available under the Apache 2.0 license, Mochi 1 allows users to generate high-quality videos directly from text prompts, rivaling industry giants like Runway’s Gen-3 Alpha and Luma AI’s Dream Machine.
Mochi 1 sets itself apart by offering free access to the model weights and code on Hugging Face, allowing users to explore cutting-edge AI video generation capabilities. However, those interested in operating the model locally will need powerful GPUs—specifically **at least 4 Nvidia H100 GPUs**.
Unlike most proprietary models, Mochi 1 is open for experimentation with a hosted playground ongenmo.ai/play. The model supports 480p resolution today, with Mochi 1 HD—a higher-definition version—set to launch later this year.
Performance and Innovation
Mochi 1 is built on Genmo’s unique Asymmetric Diffusion Transformer (AsymmDiT) architecture, with 10 billion parameters—the largest open-source video model to date. Genmo’s approach focuses on enhancing video fidelity and ensuring precision in motion, characters, and settings. As a result, it offers high-quality motion and strong adherence to user prompts, setting a new benchmark for open video generation.
Despite its prowess, Mochi 1 does have some limitations in terms of resolution and handling complex motion. Genmo has acknowledged these issues and promises further refinement in the HD version.
Backed by $28.4 Million in Series A Funding
As part of its announcement, Genmo also shared news of its successful $28.4 million Series A funding round, led by NEA and joined by several high-profile investors. This funding is set to drive Genmo’s ongoing research and development, particularly in creating **long, fluid, high-quality video generation.
The Future of AI Video Generation
Looking ahead, Genmo has ambitious plans to integrate image-to-video synthesis and further enhance user control over video outputs. Their vision extends beyond entertainment, with applications in robotics, autonomous systems, and embodied AI.
Genmo’s CEO, Paras Jain, envisions a future where AI video generation is democratized to the extent that anyone, anywhere, can create and share high-quality videos with ease.
