A two-way voice assistant in English and Chinese that does speech recognition and speech output on a local edge box. You can interrupt it mid-reply, and the language model runs either in a cloud you choose or on the device. It suits service counters, exhibition stalls, robots and meeting or hotel rooms.
RetailRobotics and educationHomes and buildings
Results reported by the design's authors on their own test setups. Yours depend on the site, so we confirm them in a pilot.
The problem
People talking to a voice terminal need to cut off a wrong answer and ask again, with replies coming from the device.
The flow
Deployment architecture
The devices on site, what carries their data, where the decision is made and where the result goes.
01SenseOn site
reSpeaker XVF3800 USB 4-Mic Array1 per terminal02DecideRuns the model
reComputer Super J30111 per terminal
reComputer Super J40121 per terminal03Your systemsWhere results land
04ActWhat happens next
How it works
The hardware
Items we stock can go straight in the bag. Everything else we source and quote with the installation.
reComputer Super J3011Jetson voice host with cloud model
reComputer Super J4012fully local voice host with Qwen3.5-4BreSpeaker XVF3800 USB 4-Mic Arrayfar-field mic with hardware echo cancellationBefore you install
Good to know
How we deliver it
Questions
Yes on Orin NX, or RK3588 with an RK1828 card, after a first online start. The cloud preset needs constant internet.
The RK3576 runs the full local speech stack and uses a cloud model for replies.
Yes. Talking during playback stops the reply, and this can be turned off.
Tell us the site and the problem. We come back with a pilot plan and a priced hardware list.
More in Acoustic & Audio AI
Based on a design published by Seeed Studio, written up by us. Symbiotes Supply is an independent reseller; product names belong to their makers.
