Microsoft Foundry expands video AI with fal

Introduction

Microsoft is expanding Microsoft Foundry’s generative AI capabilities through a new partnership with fal, adding multiple video generation models in public preview. For Azure developers and enterprise IT teams, this matters because video AI workloads often need different models for speed, quality, prompt accuracy, and cost rather than a one-size-fits-all approach.

What’s new

Microsoft Foundry now includes a new portfolio of fal-served video models as a first-party Azure service:

• Kling v3 Pro
• LTX-2.5 Fast
• LTX-2.5 Pro
• MiniMax H3 Max
• MiniMax H3 Max Turbo

These models support text-to-video and image-to-video scenarios, giving teams more options for building multimodal workflows inside enterprise applications.

Model highlights

• Kling v3 Pro focuses on cinematic output, fluid motion, native audio generation, and custom element support.
• LTX-2.5 Fast is designed for rapid iteration and lower turnaround times.
• LTX-2.5 Pro targets higher-fidelity final output.
• MiniMax H3 Max and H3 Max Turbo offer additional choices optimized for prompt adherence, aesthetics, and inference performance.

Why it matters for IT and development teams

This release is important because enterprise video AI projects typically require more than just model access. Teams also need:

• Model evaluation for quality, latency, and cost
• Production readiness features such as governance, observability, and infrastructure support
• Data residency awareness for regulated or location-sensitive workloads

Microsoft noted that the initial portfolio is hosted in the US, and Foundry is the first provider offering a US-hosted Kling endpoint. That could be a key factor for organizations with residency or compliance requirements.

Pricing considerations

Pricing varies by model and resolution, which gives teams room to align output quality with budget:

• MiniMax H3 Max Turbo starts at $0.025 per second for 480p
• MiniMax H3 Max starts at $0.05 per second for 480p
• LTX-2.5 Fast ranges from $0.09 to $0.30 per second depending on resolution
• LTX-2.5 Pro starts at $0.12 per second for 720p
• Kling v3 Pro pricing varies by audio configuration, starting at $0.1232 per second

Microsoft says Azure pricing page updates are coming in the next few weeks.

Next steps

Admins, architects, and developers should:

1. Review the new models in Microsoft Foundry preview.
2. Compare models based on quality, speed, hosting location, and cost.
3. Validate data processing and storage terms before production use.
4. Watch for updated Azure pricing details and broader model availability.

Organizations interested in scaling video AI can also look at Microsoft’s Ignite session on using fal with Microsoft Foundry for deeper implementation guidance.