SayGM (SN28) Adds MiniMax H3 for Text-to-Video and Image Animation

SayGM has added video generation to its AI gateway, allowing users to turn a text prompt, a single photo, or a pair of starting and ending frames into a short video with sound.

The feature runs on MiniMax H3, delivered through KubeTEE AI (a Bittensor subnet), bringing video creation into the platform’s existing dashboard and API.

Video is live on SayGM 🧵

A child's crayon dragon takes flight. One sentence becomes a stop-motion sunrise. Two still images become a single clip.

Here's how each one was made, so you can make your own. pic.twitter.com/jZUyFWowmY

— SayGM (@say_gm_) October 14, 2026

Inside the MiniMax H3 Integration

MiniMax H3 generates video and audio together, producing clips of up to 15 seconds at resolutions of up to 2K. Its support for text prompts, image-guided generation and transitions between two supplied frames gives users several ways to shape how a scene develops.

The integration connects two Bittensor subnets with distinct roles. SayGM operates Subnet 28, providing the gateway through which users access AI models, while KubeTEE operates Subnet 90 and supplies the infrastructure used to serve MiniMax H3.

KubeTEE builds around hardware-based Trusted Execution Environments, which can help protect workloads during processing.

The precise protections depend on the deployment and execution route, but the partnership gives SayGM another model to offer through its existing service.

How to Generate a Video on SayGM

Users can access the feature by opening the SayGM dashboard, selecting the Videos section and choosing MiniMax H3. From there, they can describe a scene in text, upload a photo to animate or provide starting and ending frames to guide the transition.

Developers can also generate videos through SayGM’s API, making it possible to add video creation to existing applications without building a separate integration with the model provider. 

One Gateway With More Creative Opportunities 

For users already working through SayGM, the addition brings video generation into the same platform they use to access other AI models.

They can move from text and static images to short video clips through a familiar interface, while developers can add video generation to applications through the existing API.

The subnet partnership also demonstrates how different parts of Bittensor can contribute to a single service.

SayGM handles model access and routing, while KubeTEE supplies the infrastructure behind the integration, connecting their capabilities in a product users can directly access.

With MiniMax H3 now available, creators and developers have another way to turn ideas and images into moving content without building a separate workflow around a dedicated video service.

SayGM’s gateway can now take a prompt or a set of images one step further and return a finished clip with sound.

➛ Start creating through SayGM.

➛ To read more about SayGM’s multi-model interface and image generation, see HERE.