
World’s first direct speech-to-intent spoken language understanding system: Fluent.ai's Automatic Intent Recognition engine ("Fluent.ai Air") offers fully offline, faster and more accurate speech understanding than conventional technologies.

Offer your users the ability to interact with their devices even without an active internet connection. Natural voice interaction in a user’s native language is driven by Fluent.ai’s ability to understand any language or combination of languages. Users can have peace of mind knowing that their private data never leaves their device.

Enable unique user interaction on wearable devices with Fluent.ai’s low memory and processing requirements. State of the art machine learning algorithms provide high accuracy in any language or accent, and even with kids’ voices and outdoor noisy environments.

High-noise environments in factories pose challenges that limit the performance and reliability of conventional speech understanding technologies. Fluent.ai technology enables robust, high accuracy speech understanding controls and solutions for noisy factory environments.
Fluent.ai Air offers OEMs and device manufacturers our on-device speech-to-intent technology for developing voice user interfaces. Our patented speech-to-intent technology directly understands a person’s speech without converting it to text first. This groundbreaking approach results in a faster, lighter, more accurate and robust speech understanding system than conventional technologies. Furthermore, it enables previously infeasible use cases and user interaction scenarios such as recognizing multiple languages concurrently and learning from users’ preferences and feedback.
Minimum Device Requirements*
Solution |
Fluent.ai AIR for MCPUs and DSPs |
Multi-WakeWord + Voice Control Engine |
---|---|---|
|
|
|
*Note that these numbers are estimated on a particular CortexM4 platform. The exact footprint might change depending on your hardware, its features, and support of various Fluent libraries for the hardware.
Minimum Device Requirements*
Vocabulary Size
-
Fluent.ai AIR for MCPUs and DSPsUp to 1000s of commands
-
Multi-WakeWord + Voice Control Engine3 - 4 WakeWords + 10 - 100 commands
RAM
-
Fluent.ai AIR for MCPUs and DSPs100 KB
-
Multi-WakeWord + Voice Control Engine128 KB
Storage
-
Fluent.ai AIR for MCPUs and DSPs500 KB
-
Multi-WakeWord + Voice Control Engine600 KB
Latency
-
Fluent.ai AIR for MCPUs and DSPs< 200 ms
-
Multi-WakeWord + Voice Control Engine200 ms
Minimum Frequency
-
Fluent.ai AIR for MCPUs and DSPs48 MHz
-
Multi-WakeWord + Voice Control Engine48 MHz
Recommended Platforms
-
Fluent.ai AIR for MCPUs and DSPsARM Cortex M4 or higher
-
Multi-WakeWord + Voice Control EngineARM Cortex M4 or higher
*Note that these numbers are estimated on a particular CortexM4 platform. The exact footprint might change depending on your hardware, its features, and the support of various Fluent libraries for the hardware.
Universal Language & Accent Support
Fluent.ai’s unique speech-to-intent technology allows us to quickly develop speech understanding models in any language and for any accent on demand.

En NA
En UK
En AUS
Sp MEX
Sp ESP
Fr FRN
Pr BRA
De
Ko
It
Mn
Hr
Pl
Ru
Tr
Features
Fluent.ai is a leader in low-power speech understanding solutions. Our systems are highly portable and require a very light footprint. We support a wide range of hardware platforms. In particular, both our WakeWord and Air systems are able to run on ARM CortexTM M4 series of microcontrollers (without needing a Floating Point Compute Unit) from AmbiqMicroTM and STMicroelectronicsTM, ARM CortexTM M7 series, ARM Cortex A series of processors, and many digital signal processors (DSPs). We have also ported our algorithms to various AI chips with specific neural network accelerators. Contact us if you have specific hardware requirements.
Supported
Hardware
Enhance your devices with Fluent.ai's
offline, robust and multilingual voice AI engine
