We’re excited to announce that Async Pro v1.0, Async’s flagship multilingual text-to-speech model, is now available on fal.
Async Pro is designed for developers building AI voice agents, customer support systems, conversational applications, content creation tools, and localization workflows. The model combines natural speech generation, multilingual support, and advanced text normalization to deliver reliable voice experiences in production environments.
With Async Pro now available on fal, developers can integrate high-quality speech synthesis into their applications while leveraging fal’s scalable inference infrastructure.
Multilingual Speech Generation
Async Pro v1.0 currently supports six languages (more languages to come in near future):
- English
- German
- French
- Spanish
- Italian
- Portuguese
The model is built on a unified multilingual architecture, allowing developers to support multiple languages through a single integration. This simplifies deployment and enables teams to expand voice experiences to new markets without managing separate models.
Async Pro focuses on maintaining consistent speech quality across languages, making it suitable for both customer-facing applications and large-scale content generation workflows.
Natural and Expressive Voice Quality
A great TTS model should sound natural, engaging, and easy to listen to.
Async Pro is optimized to generate speech with:
- Natural pacing
- Human-like intonation
- Clear pronunciation
- Smooth sentence transitions
- Consistent audio quality
The model performs well across a variety of use cases, from short conversational responses for voice agents to longer-form narration and content generation.
Advanced Text Normalization
One of the biggest challenges in production TTS is not voice quality, but correctly interpreting real-world text.
Applications frequently contain:
- Dates and times
- Currency amounts
- Phone numbers
- Measurements
- Addresses
- Abbreviations
- Mixed text and numeric content
Many TTS systems struggle with these cases, leading to unnatural pronunciations and requiring developers to build custom preprocessing pipelines.
Async Pro includes an advanced text normalization system that automatically converts written text into natural spoken language during synthesis.
For example:
- “$1,234.56” becomes “one thousand two hundred thirty-four dollars and fifty-six cents”
- “10:30 AM” becomes “ten thirty a.m.”
- “25 km” becomes “twenty-five kilometers”
This capability helps reduce pronunciation errors, improves conversational quality, and allows developers to send natural written text directly to the model without extensive preprocessing.
We recently benchmarked Async Pro’s normalization performance against leading TTS providers across a broad set of real-world scenarios. The benchmark includes dates, currencies, measurements, abbreviations, and structured text commonly found in voice agent and customer support applications.
Benchmark results are available here.
Built for Voice AI Applications
Async Pro is particularly well-suited for modern voice-first products, including:
- AI voice agents
- Customer support automation
- Conversational AI platforms
- Virtual assistants
- Content creation tools
- Audiobook and narration workflows
- Localization and multilingual applications
By combining natural speech generation with strong text understanding, Async Pro helps developers create voice experiences that sound more natural and require less manual tuning.
Why Async Pro on fal?
fal provides developers with a powerful platform for deploying and scaling AI models in production.
By bringing Async Pro to fal, developers gain access to:
- High-quality multilingual speech synthesis
- Advanced text normalization
- Simple API integration
- Managed inference infrastructure
- Scalable production deployments
This combination makes it easier to move from prototype to production while maintaining a high-quality user experience.
Get Started
Async Pro v1.0 is available today on fal.
Whether you’re building voice agents, customer support systems, content generation pipelines, or multilingual applications, Async Pro provides a simple way to generate natural, reliable speech across six languages.
We’re excited to introduce Async Pro to the fal ecosystem and look forward to seeing what developers build with it.
