The Talking Letter_(EN) - Grove Shield for Seeed Studio XIAO - dumpling—li - CodeCraft Hub
Scene: Smart Home. Edge AI application on CodeCraft Hub for Smart Home use cases.
Devices / BOM: Grove Shield for Seeed Studio XIAO, Grove Vision AI V2, Seeed Studio XIAO ESP32S3, Grove - VART Crossover Cable 4Pin 20cm (5pcs), MP3 Module V4.0, UART Interface, mini MP3 Player/Walkman, powered by WT2605X, 2mm plywood or acrylic, R型膨胀铆钉, M3*60双通六角隔离柱
Smart Home application on CodeCraft Hub. A Smart Letter Based on Edge Vision Recognition. A smart letter using Seeed Studio XIAO ESP32-S3 and Vision AI V2. It plays a recording when recognizing the target person, and a humorous response for strangers. With 'music stops when people leave,' it gives words a new form of emotional expression.
Documentation
This project is inspired by the film Grandma's Love Letter – the kind of personal affection carried by handwritten letters, which today gets buried under fragmented messages in the age of information overload. Using Seeed Studio XIAO ESP32-S3 and Vision AI V2 edge vision technology, we have created a smart letter with "eyes and a mouth." It plays a dedicated recording only when it recognizes the target person, delivering care from family and friends. When it encounters a stranger, it responds with a humorous remark, and it features "sound stops when people leave." No data is uploaded, ensuring complete privacy. With this, we hope to keep the most important people and their heartfelt thoughts forever by our side – in a way that can be both touched and heard.
Deploy steps
- Burn the model
This is an image classification vision model trained using the Vision AI V2, with three categories: me, other people, and no one.
If you want to customize your own visual recognition model, you need to retrain it.
Deploy an AI model to your device (in-app tool on CodeCraft Hub).
- Hardware Connection
Insert the XIAO ESP32-S3 into the Grove Shield: Ensure the USB port aligns with the shield's orientation and press the pins firmly into place.
Connect the XIAO ESP32-S3 to your computer via a USB Type-C cable. Note: The Type-C cable must be a data cable, not a charging-only cable.
Connect the Vision AI V2 module (visual input) to the XIAO-ESP32's SDA/SCL pins.
GND → XIAO GND
VCC → XIAO 3.3V
SDA → XIAO D4
SCL → XIAO D5
Insert a FAT32‑formatted memory card, connect a speaker, then connect to the XIAO board's serial port. TX and RX are crossed (not side-by-side)!!!!
GND → XIAO GND
VCC → XIAO 3.3V
RX → XIAO D6
TX → XIAO D7
- Firmware Flashing
This application supports the XIAO ESP32‑S3 . Select the corresponding firmware based on your board.
Click the "Flash Firmware" button below, follow the prompts to select the appropriate serial port, and connect the device. The flash address is 0x0 (default configuration).
If the device is not recognized, double‑click the Reset button on the XIAO to enter Download Mode, then try again.
Flash firmware to your device (in-app tool on CodeCraft Hub).
- Powering On and Usage
After flashing is complete, press the reset button on the XIAO‑ESP32‑S3. The system will enter "Standby Sensing" mode.
Use the Vision AI V2 to train an image classification model. You can define three image categories:
You (the target person)
Other people (non-target individuals)
Environment (no one present – simply take photos facing an empty room)
Prepare a FAT32-formatted memory card. Save two audio recordings to the root directory as 0001.mp3 and 0002.mp3 (or 0001.wav and 0002.wav).
When the target object faces the camera, the speaker plays 0001 (.mp3/.mav).
When another person faces the camera, the speaker plays 0002 (.mp3/.mav).
If no one is present, the speaker remains silent.
点击链接，查看（抖音）作品视频：https://v.douyin.com/a4bxairnIwg/
- Future Expandable Directions
Sequential Storytelling : Continuously playing a progressively advancing story based on how long the target individual stays in front of the device.
Offline Voice Interaction : Leveraging the built-in microphone capabilities of the XIAO ESP32-S3 Sense. Once the target individual hears the letter's voice, they can speak a specific secret passcode to trigger a second hidden recording, achieving a two-way, "over-the-air" conversation.
- File
0001.mp3
0002.mp3
TheTalkingLetter_2.dxf