Microsoft AI Models Reach Vercel AI Gateway With Voice, Transcription and Image Tools
Vercel AI Gateway now supports Microsoft AI models for speech generation, transcription and image creation, with documented prices and onboarding guidance.
Vercel has expanded its AI Gateway catalog with Microsoft AI, or MAI, models for speech generation, transcription and image creation. Announced on October 1, 2026, the integration gives developers a single Vercel pathway to use several MAI model families in existing projects, with published prices and documented onboarding. For teams building AI-enabled products, the practical value is access to multiple media capabilities through the same gateway and deployment workflows they already use.
According to Vercel's official announcement, the initial MAI availability includes MAI-Voice-2.1 and MAI-Voice-2.1-Flash, MAI-Transcribe-2-Streaming, MAI-Transcribe-2, MAI-Image-2.6, and related variants. The release is a broader catalog expansion rather than a single new endpoint: it brings audio input, audio output and image generation capabilities into Vercel AI Gateway.
Vercel also highlights MAI's zero data retention posture, alongside its emphasis on data control and transparency around training usage. That detail may be important for teams assessing whether an AI provider's data handling approach fits the information they intend to process. It does not remove the need to evaluate what data an application sends to a model or to design workflows around appropriate internal controls.
What Microsoft AI adds to Vercel AI Gateway
The models address distinct parts of a multimodal application experience. MAI-Voice-2.1 and its Flash variant provide expressive speech in multiple languages. MAI-Transcribe-2-Streaming is designed for live transcripts, while MAI-Transcribe-2 handles non-streaming transcription. MAI-Image-2.6 is a diffusion-based image model with image editing capabilities.
This range means a product team can consider voice, transcription and image workflows without having to treat each modality as a separate platform integration. For example, an application could use transcription for recorded or live audio, generate spoken output where it is useful, or add image generation and editing to a content-oriented workflow. The right implementation still depends on the product's data, interface and cost requirements.
Published pricing provides a starting point for cost planning
Vercel's AI Gateway catalog pages publish model-specific pricing. The units differ by model type, so teams should estimate costs using the unit that matches their intended volume, such as characters, images or hours of audio.
| MAI model | Primary capability described by Vercel | Published price |
|---|---|---|
| MAI-Voice-2.1 | Expressive speech in multiple languages | $22 per million characters |
| MAI-Image-2.6 | Diffusion-based image generation and editing | $0.01 per image |
| MAI-Transcribe-2-Streaming | Live transcription | $0.54 per hour |
| MAI-Transcribe-2 | Non-streaming transcription | $0.10 per hour |
| MAI-Voice-2 | Voice model | $22 per million characters |
The table is useful for early budgeting, but it is not a full project cost estimate. A live transcription feature, for instance, is priced per hour, whereas generated speech is priced per character and images are priced per output. Teams should therefore measure the expected activity in their own workflow before committing to a user-facing feature or internal process.
A practical route from account setup to application integration
Vercel's MAI catalog pages point developers to the basic sequence: obtain an API key, consult the documentation, call the relevant endpoint through AI Gateway, and deploy through normal Vercel workflows. That is a practical starting point for existing Vercel projects because the model call can be integrated into the same application environment instead of requiring a separate deployment process for each modality.
For a focused first implementation, teams can work through a short sequence:
- Choose one bounded use case, such as transcribing uploaded recordings or generating an image for a defined application feature.
- Match the model to the interaction: use the streaming transcription option when live transcripts are required, and the non-streaming model for audio that can be processed after capture.
- Estimate usage in the published billing unit before exposing the feature broadly.
- Integrate through AI Gateway using the API key and documentation, then deploy through the project's usual Vercel workflow.
- Review the data sent to the model and assess how MAI's zero data retention posture relates to the application's own information-handling needs.
The release does not mean every application needs all three modalities. A service that only needs accurate audio conversion can begin with transcription. A product with an established text experience may find spoken output more relevant. Image generation and editing may be most useful where visual assets are already part of the workflow. Starting with one measurable use case helps teams test user value and operating cost before adding further AI capabilities.
For businesses, the main opportunity is not simply a larger list of models. It is the option to connect a relevant AI capability to an existing web product through a developer platform that already handles the project's deployment workflow. Vercel has published availability, pricing and integration guidance, but application quality will still depend on thoughtful product design, testing and monitoring of real usage.
If your team wants to turn voice, transcription or image features into a reliable business workflow, Scalevise can help map the use case, connect the required systems and reduce manual handoffs. Our AI workflow automation services focus on practical implementations that fit existing operations, rather than adding AI for its own sake. Start by discussing an AI automation project with Scalevise.
Frequently Asked Questions
What Microsoft AI models are available through Vercel AI Gateway?
Vercel lists MAI-Voice-2.1, MAI-Voice-2.1-Flash, MAI-Transcribe-2-Streaming, MAI-Transcribe-2, MAI-Image-2.6 and related variants on AI Gateway.
What is the difference between MAI-Transcribe-2 and MAI-Transcribe-2-Streaming?
MAI-Transcribe-2-Streaming is intended for live transcripts. MAI-Transcribe-2 is the non-streaming transcription option.
How much do MAI models cost on Vercel AI Gateway?
Published examples include $22 per million characters for MAI-Voice-2.1, $0.01 per image for MAI-Image-2.6, $0.54 per hour for streaming transcription and $0.10 per hour for non-streaming transcription.
How do developers start using MAI on Vercel?
Developers can obtain an API key, use Vercel's documentation to call MAI endpoints through AI Gateway, integrate them into an existing project and deploy through usual Vercel workflows.
Conclusion
Microsoft AI's arrival on Vercel AI Gateway gives developers documented access to voice, transcription and image models through one platform. With published pricing, onboarding guidance and a stated zero data retention posture, the release gives teams a concrete basis for evaluating focused multimodal features in their existing Vercel applications.