Compact Voice / V2 preview
Big voice features.
A small place on your desk.
All the flexibility of Voice DSP Board, neatly packaged with CoreZero S3 and two small stereo speakers. A voice companion for Home Assistant, or a USB sound card for AI conversations, dictation and calls.
Built for speech first. Add external speakers when you want more from your music.
Preorder: €119.90 €109.90. Dispatch schedule to be confirmed. Not for immediate dispatch.
01 / Home Assistant
Your assistant, without another Pi.
The separate ESP32-S3 controller is included inside the enclosure. The planned pack ships with ESPHome preloaded: connect power, configure Wi-Fi and pair it with your existing Home Assistant server.
Your Assist pipeline can use local speech services or cloud services. The choice is yours; this is a satellite, not a replacement for the Home Assistant server.
02 / Simple USB audio
Connect. Select. Talk.
Use compatible USB firmware to expose a processed microphone and playback device to Windows, macOS or Linux. Select it in your voice application, conferencing software or recorder.
Install supported DSP releases through the browser-based USB Web Updater. AI apps, accounts and subscriptions are not included. Browser support and hardware revision must match the installer.
Dedicated voice processing
Less noise. More of your voice.
Four microphones + XVF3800 DSP
A dedicated audio processor handles the microphone array before audio reaches your assistant or computer. The integrated microphones form a 44 × 44 mm square.
- Beamforming: focus the microphone array on the person speaking.
- Acoustic echo cancellation: use the playback signal as a reference to reduce the sound of the device’s own speaker in the microphone stream.
- Noise suppression: reduce steady background sounds such as a fan.
- Dereverberation: reduce the effect of room reflections.
- Automatic gain control: keep speech levels more consistent.
- Direction of arrival: use the estimated voice direction for visual feedback or your application.
A wave. A press. A little light.
- 24 addressable RGB LEDs for sound direction, assistant activity and custom status feedback.
- Four physical buttons with software-defined actions.
- PAJ7620U2 gesture sensor for configurable touch-free actions such as volume or playback control.
- 3.3 V Qwiic I2C expansion via a JST 1 mm connector for compatible integrations.
Gesture mappings, button actions and LED behaviour depend on the installed firmware. Gesture control is not room-presence detection.
The XVF3800 processes audio, not language. Wake words run on the ESP32-S3, Raspberry Pi or host. Speech recognition, answers and speech synthesis belong to your chosen application. Performance depends on the room, placement, playback level and software.
Quiet replies. Bigger sound.
Voice on its own.
More sound when you want it.
Two small speakers included
The built-in stereo speakers use the 2 × 2 W output path. Think of the sound and output of a good modern smartphone: clear spoken replies and occasional music, not a substitute for a larger music speaker.
Choose your own playback system
Connect suitable passive speakers to the powerful 2 × 25 W amplifier outputs, or feed compatible powered speakers or your stereo through the 3.5 mm output. External speakers, stereo and high-power supply are not included.
Maximum amplifier capability requires a suitable 20 V supply and 4-ohm loads; it is not available from a normal computer USB port.
In the planned pack
- Voice DSP+ V2 main board with four integrated microphones.
- CoreZero S3 ESP32-S3 controller.
- Compact Voice slim 3D-printed enclosure.
- Two small stereo voice speakers.
- 24 RGB LEDs, gesture sensor and four buttons integrated on the main board.
- ESPHome preloaded as the intended shipping configuration.
Bring your own
- Home Assistant server and configured Assist pipeline for Home Assistant use.
- Compatible power source and required connection cable.
- External passive speakers or powered stereo, if desired.
A Raspberry Pi and USB-C PD charger are not included. External LINE/SQUARE microphone modules are optional, sold separately.
Connections and technical overview
| Audio processor | XMOS XVF3800 dedicated four-microphone voice DSP |
|---|---|
| Controller included | CoreZero S3: ESP32-S3, 16 MB flash, 8 MB PSRAM, 2.4 GHz Wi-Fi |
| High-power outputs | 2 × 25 W amplifier capability at 4 ohms / 20 V with suitable supply and drivers. Actual continuous output is limited by the complete power and thermal design. |
| Small speaker outputs | Separate 2 × 2 W stereo output path |
| Jack | 3.5 mm stereo audio output with connection detection and software-controlled routing |
| Computer mode | USB microphone + playback device with compatible USB firmware |
| Advanced integration | Raspberry Pi 40-pin interface, I2S audio, I2C control and SPI integration. A fitted header and revision-specific software are required. |
| User controls | 24 RGB LEDs, four buttons, PAJ7620U2 gesture sensor |
| External microphones | Optional LINE or SQUARE four-microphone array via FPC24. Internal or external array active at a time; matching firmware required. |
Hear the difference
Real audio, not just a feature list.
Echo cancellation
Compare playback echo with AEC disabled and enabled.
Noise suppression
A speech recording with fan-like background noise, before and after processing.
Far-field wake word
Local Hey Jarvis detection on a Raspberry Pi, not on the XVF3800.
These existing Voice DSP+ demonstrations illustrate the processing approach. They are not measurements or qualification of the new V2 enclosure or extension arrays.
Firmware and ownership
Your hardware. Your choice of software.
Use Home Assistant with your preferred Assist pipeline, or choose USB sound-card mode for a computer. The included CoreZero S3 connects over Wi-Fi and runs the voice-satellite application; the XVF3800 takes care of the microphone audio front end. No Raspberry Pi is needed inside the device.
Two firmware roles: the DSP USB loader updates compatible standalone audio firmware. It does not install the Home Assistant application on the ESP32-S3. Use the controller installer for that application.
Choose firmware for the exact hardware revision. Current V1 images are not V2 updates. The V2 software integration is in progress; the planned shipping setup includes ESPHome preloaded. Wi-Fi pairing and your Home Assistant server configuration are still required.
Questions before you choose
Can it work without Home Assistant?
Yes, in supported USB audio mode it can be a microphone and playback device for your computer. Your computer runs the application.
Does the DSP run ChatGPT or a wake word?
No. The DSP cleans up audio. Wake-word and application processing run on the controller or host; AI services run wherever your selected application provides them.
Can I use it for music?
Yes, occasionally. For fuller music playback, add external speakers or choose the larger Full Speaker format.
