Built for real-world Edge AI

Explore how femtoAI enables ultra-efficient, always-on intelligence across leading device categories.

Ear Buds
True Wireless Stereo (TWS)
Clearest call quality and sound detection for dynamic system-level optimization
The Challenge
True wireless earbuds demand always-on AI audio processing for premium sound quality while operating under extreme constraints on power, latency, and size.
Why femtoAI
High-quality noise reduction for in-person conversations and voice enhancement during calls

Multimodal AI using multi-sensor data fusion to unlock industry-leading noise suppression

Ultra-low latency (~4 ms) processing for natural, face-to-face speech enhancement

Ultra-low power wake-word detection and voice commands (~250µW)

Private, on-device AI with compressed models under 1MB and no cloud dependency
See It In Action
Clara Speech Enhancement
Multimodal Call Denoising
Clara Power Measurement
Clara AI Speech Enhancement running on SPU-001 with background highway noise.
Presented by Scott Reid, Chief AI Officer.

Multimodal Call Denoising combining microphone and VPU input
Presented by Roja de Cande, Director of Product Management.

Clara AI Speech Enhancement with live spectrogram and power measurement, running at just 1.2mW of power for real-time processing.
Presented by Bill Hoppin, Chief Business Officer.
Ultra-low latency
4ms

Ultra-low power
< 1mW

Small memory footprint
< 1MB
Smart Glasses
Smart Glasses
Natural voice interaction and context-aware sensing for all-day, always-on intelligence
The Challenge
Smart glasses must continuously process audio, language, and vision intelligence while meeting strict limits on power, thermal budget, and form factor.
Why femtoAI
Ultra-low-power wake-word detection enabling the main SoC to remain in deep sleep until needed

High-quality noise reduction for calls and face-to-face voice boost

Feedback cancellation optimized for open-fit acoustic designs

Multimodal AI combining audio and sensor data for robust context awareness

Private, on-device AI with compressed models under 1MB and no cloud dependency
See It In Action
Alexa Wake Word
Person Detection
Multimodal Call Denoising
Alexa wake word detection running on SPU-001 with live power measurements.
Presented by Pong Trairatvorakul, Director of Strategy.

Person detection Visual Wake Word application running on SPU-150.
Presented by Roja de Cande, Director of Product Management.

Multimodal Call Denoising combining microphone and VPU input
Presented by Roja de Cande, Director of Product Management.
Always-on sensing

All-day wake word
250 microwatts

Low latency
On-device intelligence

Small model footprint
< 1MB
Hearing Aid
Hearing Aid
High-fidelity, low-latency, low power AI for all-day wear
The Challenge
Hearing aids must adapt to complex, noisy real-world environments while delivering all-day battery life in extremely small form factors with limited power and thermal headroom.
Why femtoAI
Adaptive, AI noise reduction that suppresses both steady-state background noise and transient acoustic events

Ultra-low power operation enabling all-day AI processing on a single charge

Ultra-low latency (4ms) preserving natural sound and face-to-face speech cues

Production-ready audio pipeline, including WDRC, feedback cancellation, beamforming, dynamic EQ, and sound scene classification
See It In Action
Clara Speech Enhancement
Clara Power Measurement
Clara AI Speech Enhancement running on SPU-001 with background highway noise.
Presented by Scott Reid, Chief AI Officer.

Clara AI Speech Enhancement with live spectrogram and power measurement, running at just 1.2mW of power for real-time processing.
Presented by Bill Hoppin, Chief Business Officer.
Clear natural speech

All-day battery life
< 1mW

Ultra-low latency
4ms
Headphones2
Headphones
Immersive audio and clear voice communication across use cases
The Challenge
Modern headphones are expected to deliver immersive sound, clear voice communication, and effective noise cancellation across calls, media, gaming, and travel—while maintaining low power, low latency, and long battery life.
Why femtoAI
High quality AI noise reduction and speech enhancement using multi-mic and multi-sensor algorithms for superior call clarity and immersive listening experiences

Low-power, always-on intelligence enabling adaptive ANC, voice activation, and sound scene awareness

Ultra-low-latency audio processing optimized for real-time applications such as calls and gaming
See It In Action
Clara Speech Enhancement
Multimodal Call Denoising
Clara Power Measurement
Clara AI Speech Enhancement running on SPU-001 with background highway noise.
Presented by Scott Reid, Chief AI Officer.

Multimodal Call Denoising combining microphone and VPU input
Presented by Roja de Cande, Director of Product Management.

Clara AI Speech Enhancement with live spectrogram and power measurement, running at just 1.2mW of power for real-time processing.
Presented by Bill Hoppin, Chief Business Officer.
Multi-mic fusion
For optimal sound quality

Ultra-low latency
4ms
Speakers
Smart Speakers
Always-on voice interaction with privacy and low power
The Challenge
Smart speakers must deliver far-field voice recognition, natural language interaction, and high-quality audio playback in acoustically diverse home environments. These systems operate continuously, requiring always-on listening while balancing power efficiency, latency, cost, and thermal constraints. Privacy concerns and reliance on cloud connectivity further raise expectations for responsive, on-device intelligence.
Why femtoAI
Ultra-low power consumption for always-on, far-field voice detection and wake-word recognition

AI noise reduction and echo cancellation for accurate voice capture in reverberant and noisy environments

Low-latency spoken language understanding for responsive and natural user interactions

On-device and private intelligence that reduces dependence on cloud processing
See It In Action
Smart Speaker
Alexa Wake Word
Smart speaker voice command with acoustic echo cancellation (AEC), demoed on the Cear Pave.
Presented by Corey McCall, Principal Electronics Engineer.

Alexa wake word detection running on SPU-001 with live power measurements.
Presented by Pong Trairatvorakul, Director of Strategy.
Instant response
< 20ms

Private & secure
On-device intelligence

Ultra-low power
< 1mW
Home Appliance
Smart Home & Digital Appliance
Efficient on-device AI for responsive, private operation
The Challenge
Smart home devices and digital appliances require instant, reliable user interaction while maintaining privacy and operating under strict cost, size, and power constraints. Current systems rely on cloud connectivity, resulting in latency and compromised privacy.
Why femtoAI
Low-latency on-device inference for instant response to user commands

Compact and cost-efficient AI enabling integration into space and cost constrained appliances

Offline and private operation ensuring reliability and user trust without cloud connectivity

Low-power always-on intelligence for continuous responsiveness
See It In Action
Smart Lights with ESP32
Smart Home SLU
Smart home voice commands running on SPU-001 integrated with ESP32, controlling physical lights.
Presented by Jon Russo, Senior Solutions Architect.

Sentence level understanding for flexible smart home commands.
This demo controls a virtual smart home available on our Developer Portal.
Presented by Mathieu LeGoc, Principal Firmware Engineer.
Instant response
< 20ms

Private & secure
On-device intelligence

Ultra-low power
For always-on operation
Security Camera
Security Camera
Always-on intelligence at the edge with minimal power
The Challenge
Smart cameras and video doorbells must deliver always-on audio and vision intelligence for accurate event detection, person recognition, and situational awareness in constrained systems: power, size, thermals. Low power, high reliability, and on-device processing are increasingly critical to meet user expectations for real-time alerts and privacy.
Why femtoAI
Ultra-low-power consumption for always-on visual wake word and sound event detection

Power and accuracy enhancing host processing through real-time data gating and intelligent preprocessing

On-device and private intelligence that reduces dependence on cloud processing

Low-latency on-device inference for rapid detection, alerting, and response
See It In Action
Person Detection
Sound Event Detection
Person detection Visual Wake Word application running on SPU-150.
Presented by Roja de Cande, Director of Product Management.

Baby cry Sound Event Detection running on SPU-001 that can wake up the host processor only when needed.
Presented by Roja de Cande, Director of Product Management.
Small model footprnt
< 1MB

Ultra-low power
< 1mW

On-device intelligence
Assembly Arm
Robotics
Low-latency perception and control for autonomous systems
The Challenge
Autonomous robots require fast, reliable perception and control loops while operating within strict power budgets, especially for mobile and battery-powered platforms. Latency directly impacts safety and performance.
Why femtoAI
Ultra-low-power consumption for audio and sensor intelligence for voice interaction and situational awareness

Low-latency multimodal sensor fusion to enable fast, context-aware responses

Efficient on-device processing reducing dependence on external compute or connectivity
See It In Action
Low-latency
For realtime control loops

On-device processing
Reduce external dependence