Genmo

Genmo builds open-source video generation models like Mochi 1, with a browser playground and local customization options.

AI 동영상 생성기 텍스트-동영상 변환

웹사이트 방문

Overview

Genmo is a research lab focused on open video generation models. Its main public product is Mochi 1, an open-source text-to-video model available through a hosted playground and an open repository.

The site positions Mochi 1 as a model for generating short videos from written prompts, with emphasis on motion quality and prompt adherence. Genmo also says users can run it locally, customize it for their needs, or use ComfyUI, while the preview release is available under the Apache 2.0 license for personal and commercial use.

Core features

Text-to-video generation

Mochi 1 is Genmo’s open-source text-to-video model for turning written prompts into generated videos.

Prompt adherence

The model is designed to follow detailed prompts closely, with the blog describing strong prompt adherence for characters, settings, and actions.

Motion-focused output

Genmo says the preview generates smooth video at 30 frames per second for clips up to 5.4 seconds, with a focus on coherent motion.

Local and ComfyUI workflows

The open-source repository can be run locally, and Genmo also says Mochi can be used through ComfyUI.

Hosted playground

The product includes a hosted playground where users can test Mochi 1 directly in the browser.

Open release for development

The release is open under Apache 2.0, with open weights and source code available for developers.

Practical use cases

Text-to-video prototyping
Create short prompt-driven videos from text when you want a browser-based starting point in the playground before moving to local runs or customization.
Local experimentation
Run the open repository locally to adapt the model for your own workflow, experiments, or research work.
ComfyUI-based generation
Use ComfyUI support if you prefer a node-based creative workflow for testing Mochi 1 output.
Model evaluation and comparison
Evaluate model fit for applications that care about motion quality and prompt adherence, such as research previews or product experiments.
Licensed preview usage
Generate personal or commercial video clips under the preview’s Apache 2.0 license, while accounting for the current 480p and short-duration limits.

Pros and Cons

Pros

Open-source release with Apache 2.0 licensing for personal and commercial use.
Hosted playground makes it possible to try the model without local setup.
Local repository access supports customization and developer workflows.
The model is described as having strong prompt adherence and improved motion quality.

Cons

The initial release is limited to 480p in the preview described on the blog.
Genmo notes that animated content is not a strong fit for the current model.
The blog says minor warping and distortions can appear in some extreme-motion edge cases.

FAQ

How can I try Mochi 1?

Genmo’s playground is the easiest way to try Mochi 1 without setting up the open-source repository. The site also points to an open-source release for users who want to run it locally or customize it.

Is Mochi 1 open source?

The blog post states that the initial research preview of Mochi 1 is available for personal and commercial use under the Apache 2.0 license.

What output quality does Mochi 1 support today?

The first release is a 480p base model. The blog says Mochi 1 HD is planned for later, but it is not part of the initial release.

How do developers access the model?

Genmo says Mochi 1 can be run locally from the open-source repository and also used through ComfyUI. The blog additionally mentions open weights on HuggingFace and API partners.