Genmo Lancia Mochi 1: A new model of open source video generation | Ai image generator | Midjourney | Ai image generator free online | Turtles AI

Genmo presented Mochi 1 preview, an Open Source model of generation video based on AI, promising improvements in the quality of movement and in adherence to user instructions. With a financing of 28.4 million dollars, the company aims to develop the creativity of AI.

Key points:

• Mochi 1 is an open source model for video generation, with advanced movement capacity.
• The company collected $ 28.4 million in funding to develop creative.
• The model is able to generate fluid videos at 30 fps and supports durability of up to 5.4 seconds.
• Mochi 1 includes a new processing architecture, Asymmetric Diffusion Transformer, for efficient optimization of prompts.

Geno has announced the preview of his new video generation model, Mochi 1, available in Open Source. This tool is designed to significantly improve the quality of the movement in the videos and to ensure accurate correspondence with the indications provided by users. Contrary to other AI models, often inclined to incorrect interpretations of requests, Mochi 1 has been developed to scrupulously follow the clear and detailed instructions of users, thus resulting more reliable in the generation of video content. In addition to the model, an interactive playground was presented, accessible for free, which allows users to test Mochi 1. The weights of the model can also be found on the Hugging Face platform, a resource widely used for AI models.

Genmo also communicated that he recently obtained $ 28.4 million in a Serie A financing round, with the support of investors such as Nea, The House Fund and Gold House Ventures. These funds will be used to further explore and develop what the company calls "the right brain of the general AI", a concept associated with creativity. Mochi 1 is considered an initial step towards the construction of creative skills in AI, in a sector that has seen significant investments in the video generation, especially after the launch of solutions such as the video generators of Runway ai Inc. and Sora di Openai.

The new model establishes high standards for realistic movement in videos, incorporating physical dynamics such as the fluid movement, the simulation of fur and hair, and, in particular, human movement. Mochi 1 is capable of generating fluid videos at 30 frames per second for a maximum duration of 5.4 seconds, which represents an agreement in the current sector. When users formulate clear requests, the model proves particularly precise in the production of videos that faithfully reflect the instructions provided, allowing detailed control over characters, scenes and other creative variables.

To develop Mochi 1, Genmo has adopted a diffusion model containing 10 billion parameters, a significant figure that contributes to the precision and accuracy of the model. The Asymmetric Diffusion Transformer architecture, used by the team, allows efficient management of user and token compressed video prompts, optimizing the processing of the text in favor of visual elements. This approach creates videos through joint use of text and visual tokens, similar to what is made by Stable Diffusion 3, but with a capacity four times higher in the flow of text, thanks to a wider hidden dimension. The asymmetrical design of architecture allows to reduce memory consumption.

The preview of Mochi 1 currently offers a 480p video quality, with the intention of releasing a complete version by the end of the year, which will include Mochi 1 HD. This advanced version will support video generation at 720p, promising higher loyalty and even more fluid movement. Genmo stressed that Mochi 1 was developed from scratch, representing the largest Open Source video generation model currently available. With over 2 million users already active in the closed models of the company, the release of the source code and the Mochi 1 weights under the open source license apache 2.0 on platforms such as Github and Hugging Face represents an important opportunity for developers and researchers in the field of the AI.

Pending the complete release, Mochi 1 innovation marks an important step in the panorama of the video generation based on AI.