Smart Speaker and Alarm Clock
Introduction
Smart speakers and AI alarm clocks share one on/off-cloud voice solution. Wake word, dialogue, multi-source playback, and schedules run on the same software, and the device can ship as a speaker, bedside clock, story player, or desktop assistant. The online path uses RTC or WebSocket to a large-model or voice-cloud service. Earlier ASR + NLU designs may use MQTT or HTTP. Offline mode keeps only the wake word and local command words for home control without a network.
One SoC runs the audio front end, codecs, streaming playback, and peripherals such as Mesh, IR, motors, and a display, without a separate voice DSP. When classic Bluetooth is required, use ESP32-S31; an external voice DSP on ESP32 is also possible.
flowchart LR
Wake["Wake word"]
Chat["Cloud or on-device chat"]
Player["Multi-source playback"]
Alarm["Alarm and schedule"]
Wake --> Chat
Chat --> Player
Alarm --> Player
Applications
Home entertainment: smart speaker or network radio, playing online stations and a local library
Children’s education: story player or early-learning terminal, with wake word and cloud content
Smart home: voice assistant or voice-enabled appliance, controlling lights and devices over BLE Mesh or IR
Desk and bedside: AI alarm clock with schedule, weather, and voice reminders
Features
Wake word, plus noise suppression, acoustic echo cancellation, automatic gain control, and microphone arrays; the wake word can be customized
Multi-turn online dialogue, or offline command-word control; the online side can connect to mainstream voice-cloud and large-model services
Playback from Wi-Fi (HTTP, HLS, DLNA, AirPlay), Bluetooth, and local storage, with barge-in, resume after power loss, and fade
Alarms, schedules, and a display for weather, clock faces, or expressions
Audio codecs and EQ; IR or Mesh home control as a product extension
Hardware and software
Common SoCs for voice products are ESP32-S3, ESP32-S31, and ESP32-P4. Use ESP32-S31 when classic Bluetooth is required. Display products pick the SoC by panel interface and resolution.
The audio front end is documented in ESP-SR. Playback and format handling are in ESP Audio Codec and ESP-Audio-Effects. Real-time dialogue can follow ESP-WebRTC Real-Time Communication Solution. Example projects live in adf_examples/ai_agent.