Kling, LTX and MiniMax video models are rolling out in public preview in Microsoft Foundry, through a new partnership with fal. Microsoft is offering them as a first-party Azure service, so developers can generate video from text or from an existing image without standing up a separate video stack.
The announcement comes from Naomi Moneypenny on the Microsoft Foundry blog (November 5, 2026, updated November 6). The first models are Kling v3 Pro, LTX-2.5 Fast, LTX-2.5 Pro, MiniMax H3 Max and MiniMax H3 Max Turbo.

What each model is for
Not every video job starts the same way or aims for the same result. One team might turn a written brief into several draft clips. Another might animate an approved product photo. A third might chain image generation and video generation into one workflow. Microsoft’s point is that quality, motion, creative control, turnaround time and cost do not all peak in a single model.
The preview gives developers a small portfolio instead of one default choice. The models are served by fal and sourced from the companies that build them. Microsoft describes fal as a generative media platform with APIs for more than 1,500 models across image, video, audio, 3D and generative worlds. This launch is a selected set, with more models planned over time.
| Model | What Microsoft says it is for | Inputs listed |
|---|---|---|
| Kling v3 Pro | Cinematic visuals and fluid motion, with native audio generation and custom element support. Stronger control when a still image becomes a directed scene. | Text to video, image to video |
| LTX-2.5 Fast | The speed-optimized option for rapid drafts and previews, with synchronized audio. | Text to video, image to video |
| LTX-2.5 Pro | The quality-optimized option for higher-fidelity finished output, also with synchronized audio. | Text to video, image to video |
| MiniMax H3 Max | A post-trained MiniMax H3 variant focused on prompt adherence, aesthetics and inference performance. | Text to video, image to video, reference to video |
| MiniMax H3 Max Turbo | The faster MiniMax option, with the same post-training focus on quality and speed. | Text to video, image to video |
In their words
“fal has seen incredible adoption this year alone with generative media across all industries from advertising to marketing to gaming. Partnering with Microsoft allows both of us to truly enable the wide creative ecosystem and ensure that all AI solutions that lean on generative media, becomes a multi-model one.”
Burkay Gur, Cofounder, to Microsoft
What the preview costs
Microsoft published prices in US dollars per second of generated video. The figures vary by model and resolution. Kling is listed by audio options instead of resolution, and it uses a US data zone deployment. The LTX and MiniMax rows use Global Standard.
| Model | Deployment | Resolution or option | Price per second (USD) |
|---|---|---|---|
| MiniMax H3 Max | Global Standard | 480p | $0.05 |
| MiniMax H3 Max | Global Standard | 768p | $0.08 |
| MiniMax H3 Max | Global Standard | 1080p | $0.16 |
| MiniMax H3 Max Turbo | Global Standard | 480p | $0.025 |
| MiniMax H3 Max Turbo | Global Standard | 768p | $0.04 |
| MiniMax H3 Max Turbo | Global Standard | 1080p | $0.08 |
| LTX-2.5 Fast | Global Standard | 720p | $0.09 |
| LTX-2.5 Fast | Global Standard | 1080p | $0.13 |
| LTX-2.5 Fast | Global Standard | 1440p | $0.19 |
| LTX-2.5 Fast | Global Standard | 4K | $0.30 |
| LTX-2.5 Pro | Global Standard | 720p | $0.12 |
| LTX-2.5 Pro | Global Standard | 1080p | $0.17 |
| Kling v3 Pro | US Data Zone | Audio off | $0.1232 |
| Kling v3 Pro | US Data Zone | Audio on | $0.1848 |
| Kling v3 Pro | US Data Zone | Audio + voice control | $0.2156 |
Microsoft says these prices will be updated and listed on the Azure pricing page in the coming weeks. Treat the table as the figures published with the preview announcement, and check Azure for later changes.
What enterprise teams should check
The initial portfolio is hosted in the United States. Microsoft says Foundry is the first provider to offer a US-hosted Kling 3 endpoint. That matters if your organization has data residency rules: review each model’s documented data processing and storage terms before you send it production content.
Because this is a public preview, features, availability and pricing can still change. Microsoft’s broader pitch is that teams can evaluate the models, then use Foundry’s evaluation, optimization, governance, security, observability and infrastructure tools as they move toward production. For another recent Foundry model addition, see our note on GPT-6 Astra, Sol and Luna in Microsoft Foundry.
How to try it
The models are available to try in Microsoft Foundry. Microsoft also maintains a product overview on Azure.
Source: Microsoft.
Discover more from Windows Mode
Subscribe to get the latest posts sent to your email.
Add your first comment to this post