Microsoft and fal add five video models to Foundry in public preview

Microsoft and fal are adding five video-generation models to Microsoft Foundry in public preview, bringing fal-served options from Kling, LTX and MiniMax into the platform’s multimodal catalogue.

The initial lineup includes Kling v3 Pro, LTX v2.5 Fast, LTX v2.5 Pro, MiniMax H3 Max and MiniMax H3 Max Turbo. All five support text-to-video and image-to-video generation. MiniMax H3 Max also supports reference-to-video generation using images, video or audio, according to Microsoft’s deployment documentation.

The models span two Azure deployment options. Kling v3 Pro is available through a US Data Zone deployment, while the LTX and MiniMax models use Global Standard pay-as-you-go deployments. Microsoft says the models are sourced from their developers and served by fal.

Before deploying the models from the Foundry catalogue, customers must register the Fal.EnableDeploy preview feature for an Azure subscription. Microsoft says fal operates the services outside Microsoft facilities, and that deployment involves sharing organizational data with fal and moving it out of Microsoft.

The rollout expands Foundry’s video-generation choices after Microsoft put its multi-model Copilot harness into public preview through Foundry. Kling v3 Pro includes native audio generation and custom-element support. Microsoft describes LTX v2.5 Fast as the speed-focused option for iteration and LTX v2.5 Pro as the higher-quality option; both generate synchronized audio. fal describes H3 Max as a post-trained version of the MiniMax H3 base model.

Microsoft’s announcement and deployment documentation do not provide a general-availability date for the five models or disclose commercial terms for the partnership.