Bring Cloud-Standard Voice AI to Microcontrollers.

Voice features — now offline on MCU: ASR, TTS, and Small LM.

ASR (Automatic Speech Recognition) — IP block

  • Available: English.
  • Customization: new languages, domain-specific vocabularies, constrained grammars.

TTS (Text-to-Speech) —
 IP block

  • Available: English.
  • Customization: new languages, custom voices (tone, style, brand identity).

Small LM (Lightweight Language Models)

  • Ready IP: TinyStories (EN) — pre-trained, fully offline storytelling.
  • Custom Small LMs — tailored for intents, Q&A, dialogs, or new languages.

Customization Options

  • Add new languages and domain vocabularies.
  • Design voice flows with clarifications and confirmations.
  • Create custom TTS voices.
  • Optimize for RAM, Flash, and latency on your platform.

Platforms We Support

  • ESP32 (incl. ESP32-S3) → fast PoC and cost-sensitive devices.
  • Arm® Cortex™-M + Arm® Ethos™-U55 / other AI SoCs → high performance, energy-efficient production.
  • Migration path: start on ESP32 → scale to Arm® Ethos™-U55.

Applications

  Smart appliances

natural dialogs instead of rigid commands:
“Start cooking.” → “For how many minutes?” → TTS: “Cooking started, 10 minutes remaining.”

Toys & EdTech

interactive learning and safe offline storytelling (AI Teacher, TinyStories).

Access control 

ASR turned into offline numeric PIN (VoicePIN).

OEM Benefits

  • Faster time-to-market — PoC in weeks, not months.
  • Lower OPEX — no recurring cloud costs, fewer complaints.
  •  Minimal BOM increase  — add offline voice with negligible hardware cost.
  •  Pro SKU differentiation  — premium devices with offline voice features.
  •  De-risked adoption  — tested by the DIY community before reaching OEM scale.

DIY Proof (Social Proof)

our DIY product with three apps showing how one ASR + Intent stack enables multiple use cases:

For OEMs, the VoxControl Kit is proof that one stack can power very different devices — from toys to appliances to security systems.

An offline speech module that lets your device talk naturally without the cloud.

For OEMs, the VoxControl Kit is proof that one stack can power very different devices — from toys to appliances to security systems.

How We Work

Evaluate
Pilot (PoC)
License & Integrate
  1. Evaluate — define target sounds, access scenarios, hardware.
  2. PoC — 2–3 weeks, fixed scope, measurable results.
  3. License & Integrate — adapt IP for production; PoC fee credited.

Deliverables

  • ASR & TTS (EN) IP binaries, APIs, integration guides.
  • TinyStories (EN) ready-to-use IP.
  • Optional: custom Small LMs, language packs, TTS voices.
  • Metrics pack (latency, energy, memory).
  • Demo video on your device.

Licensing

  • PoC license — quick validation.
  • Project license — tuned models, integration support.
  •  Volume license — mass deployment.
No lock-in: portable IP, stable APIs, integration code stays with OEM.

Ready to add offline dialogs to your devices?

Start with TinyStories or request a PoC tailored to your product.
Request PoC