
1 / 4 · Click to enlarge
AI Summary
An offline voice recognition module that recognises 121 fixed commands and learns up to 17 custom words. It talks to Arduino, Raspberry Pi and ESP32 boards over I2C or UART, so voice control works without an internet connection.
Key Highlights
- Offline recognition chip with 121 built-in fixed command words
- Learns up to 17 custom commands, including whistles and snaps
- Dual microphones for better noise resistance and longer range
- I2C at 0x64 or UART communication with Gravity interface
- Onboard speaker plus external speaker connector for voice feedback
- Works with 3.3V and 5V Arduino, Raspberry Pi and ESP32 boards
Ideal For
- Makers building voice-controlled smart home devices
- Arduino and Raspberry Pi hobbyists exploring speech interfaces
- Robotics builders who want hands-free commands
- Educators teaching sensor and communication basics
- ESP32 developers adding offline voice triggers
Buying Tip
Confirm your controller can supply up to 370 mA at 5V and that I2C address 0x64 is free on your bus before wiring the module in.
Elecvora expert review
Our verdict
A well-rounded offline voice module that removes cloud dependency and offers both I2C and UART for flexible integration. The dual microphones and speaker feedback make it more practical than simple single-mic alternatives.
AI-assisted analysis based on the manufacturer's specifications. Not a hands-on lab test.
Pros
- Works fully offline with no network needed
- 121 fixed commands plus 17 trainable commands
- Dual microphones improve noise handling
- I2C and UART options suit many controllers
- Onboard and external speaker support
Cons
- Only 17 custom commands can be added
- Fixed I2C address 0x64 limits bus planning
- Up to 370 mA at 5V may challenge small supplies
Best for
Makers adding local voice control to Arduino, Raspberry Pi or ESP32 builds
Not ideal for
Projects needing large custom vocabulary or cloud-based speech recognition
Specifications
| Operating Current (mA) | ≤370 mA (5V) |
|---|---|
| Communication Mode | I2C/UART |
| I2C Address | 0x64 |
| Microphone sensitivity | -28db |
| Module Size | 49×32 mm/1.93×1.26” |
| Weight (g) | 16 |
| Shipping Weight | 0.035 kg |
| Shipping Dimensions | 8 × 7 × 3 cm |
Frequently Asked Questions
Does this module need an internet connection?
How many commands can it recognise?
Which controllers are supported?
What power does it require?
How do I connect it to my board?
Can it play back recognition results?
Overview
The DFRobot Gravity Offline Voice Recognition Sensor gives your project a local voice interface that does not depend on a cloud service or network connection. Recognition happens on the module itself, so commands are processed on the spot and no audio needs to leave the device. The board combines a compact 49 × 32 mm footprint with a Gravity connector, making it straightforward to add to existing builds.
Two microphones handle input, which helps the module cope with background noise and extends the distance over which commands can be picked up. Recognition results can be announced through the built-in speaker or through an external speaker connected to the dedicated interface.
Key Features
- Offline operation: the recognition chip works without an internet connection.
- Fixed command set: 121 commonly used command words are available out of the box.
- Self-learning: up to 17 extra commands can be trained by voice, and any sound can become a trigger.
- Flexible wiring: choose I2C at address 0x64 or UART.
- Status indicators: a red power LED and a blue recognition LED show what the module is doing.
- Audio feedback: onboard speaker and an interface for an external speaker.
Specifications Explained
The module draws up to 370 mA at 5V, so it is worth checking that your supply can deliver that current alongside the rest of your circuit. The microphone sensitivity of -28 dB describes how readily the microphones pick up sound, which pairs with the dual-mic design to support recognition in noisier rooms.
I2C address 0x64 is fixed, so plan your bus around that address if you already use other I2C devices. UART gives you an alternative when the I2C bus is busy or when you prefer a serial link. At 49 × 32 mm and 16 g, the board fits neatly into small enclosures.
Compatibility and Requirements
The sensor supports 3.3V and 5V controllers and is listed as compatible with Arduino UNO, Arduino Leonardo, Arduino MEGA, FireBeetle series boards, Raspberry Pi and ESP32. You will need a controller, a suitable power supply and a way to connect the Gravity interface to your board. Library files are available from DFRobot to help you get started.
Typical Uses
Typical projects include voice-controlled lamps and appliances, interactive toys, robotics commands and any build where a local spoken trigger is more convenient than a button or app.

Quick specs
- Operating Current (mA)
- ≤370 mA (5V)
- Communication Mode
- I2C/UART
- I2C Address
- 0x64
- Microphone sensitivity
- -28db
- Module Size
- 49×32 mm/1.93×1.26”
- Weight (g)
- 16


















