In-Store & Counter Service QA
Scenario details
For fixed counters such as retail checkouts, bank and service-hall windows, and pharmacy or clinic desks. The three tabs below cover what the counter hardware does.
Mic-array noise reduction, directional pickup across the counter
A reSpeaker mic array sits on the counter. Echo cancellation, beamforming and noise suppression run on the array's own DSP, so staff and customer are both captured within about 3 m. It outputs USB audio straight into the edge box. Rated for steady background noise at or below 70 dB.
Core Advantages
- On-array DSP noise reduction
- Directional pickup, both sides
- Plug-and-play USB audio
Docs
Edge box transcribes on site; audio stays in the store
The mic array connects to an edge box in the store, which turns speech into timestamped text sentence by sentence and uploads text only; raw audio can be discarded. One reComputer RK3576 transcribes 12 Chinese streams at once. If the back end is unreachable, transcription continues and is cached locally, then uploaded when the connection returns.
Core Advantages
- Sentence-level live text
- Audio stays in the store
- Local cache when offline
Voiceprint matching separates staff from customers
With voiceprint matching on, the edge box computes a voiceprint for every sentence and labels the speaker. Staff enroll their voiceprints on the device's local web page, so their lines are labeled separately from customers'. Voiceprint matching is a local similarity grouping, not identity verification.
Core Advantages
- Speaker label per sentence
- Staff voiceprint enrollment
- Voiceprints stay local
Deployment & Selection Guide
The configurator narrows it down to a model by transcription mode, pickup setup and compute; the two architectures and device comparisons are in the cards below.
Not sure which to choose? See the full selection guide →
Transcription Mode
Cloud needs no extra box; local keeps audio on site
Option A — Cloud transcription: reSpeaker array → Network → SenseCraft Voice → transcript reaches your business system over HTTP / MQTT
Option B — Local transcription: reSpeaker array → Edge box → Local ASR → Text upload only
Core Advantages
- Cloud: no extra edge box, works wherever there is a network
- Local: audio stays on site, only text is uploaded
- Both use: the same reSpeaker array for capture
| Image | Product | Tier | Capabilities | Measured Streams | Scale |
|---|---|---|---|---|---|
 | reRouter CM4 1432 | Entry | Single-language local transcription | Not measured | Small single-point deployments, validated reference design |
 | reComputer AI R2130-12 | Advanced | Real-time multilingual transcription | Not measured | Medium scale, latency-sensitive |
 | reComputer J4012 | High throughput | Real-time multilingual transcription | 48 streams (p95 789 ms) | Many concurrent streams on one box |
 | reComputer J3011 | High throughput | Real-time multilingual transcription | 32 streams (p95 293 ms) | Many concurrent streams on one box |
 | reComputer RK3576-30 | Standalone | Multilingual transcription + local analysis | 12 streams (p95 1067 ms) | High privacy or offline environments |
 | reComputer RK3588-30 | Standalone | Multilingual transcription + local analysis | 8 streams (p95 902 ms) | High privacy or offline environments |
CM4 for light single points, J-series for high concurrent streams
Local transcription runs on an edge box, with a reSpeaker array capturing the audio. reRouter CM4 is the official SenseCraft Voice reference hardware; confirm other models with the solutions team. The "measured streams" column is the recommended number of concurrent SenseVoice (Chinese) streams for clips up to 4 s, where the final text stays under 1.5 s at p95; timing runs from the end of speech to the final result and excludes VAD, speaker embedding, punctuation, upload and storage. Numbers are measured on the retail_voice reference design.
Core Advantages
- Light transcription: CM4 handles single small points
- High concurrency: J40 delivers 48 verified streams
- Standalone offline: RK3576 supports local offline analytics
| Image | Product | Type | Range | Coverage | Amp | Algorithms |
|---|---|---|---|---|---|---|
 | reSpeaker Lite | Linear dual-mic | ~3m | 180° | 5W | AEC · DoA |
 | reSpeaker XVF3800 | Circular quad-mic | ~5m | 360° | 5W | AEC · DoA · Multi-beam |
 | reSpeaker XVF3800 (with Case) | Circular quad-mic | ~5m | 360° | 5W | AEC · AGC · DoA · VAD · Multi-beam |
 | reSpeaker Flex Circular-4 | Circular quad-mic | ~5m | 360° | 10W | AEC · DoA · Multi-beam |
 | reSpeaker Flex Linear-4 | Linear quad-mic | ~5m | 180° | 10W | AEC · DoA · Multi-beam |
Directional noise canceling for counters, omnidirectional for meeting tables
Pick reSpeaker Lite for one-to-one counters with speakers within 3 m. Pick XVF3800 for 360° pickup at windows and consultation desks — the cased version sits on the counter, the bare board goes into your own enclosure. Pick Flex when you need a 10 W amplifier or a linear layout. For staff who wear the recorder, use reSpeaker Clip; see In-Person Conversation Intelligence.
Core Advantages
- Directional pickup ~3 m at close-range counters
- 360° omnidirectional pickup within 5 m radius on desks
- Onboard acoustics preset, ready without tuning
Frequently Asked Questions
Yes. Seeed provides ODM/OEM customization services that can integrate microphone arrays and edge AI modules into a single hardware design, suitable for customers who need unified procurement or branded deployment. If you have customization needs, feel free to reach out: Ask about ODM/OEM customization →
Yes. SenseCraft Voice provides open APIs, and transcription results can be pushed in real time through standard interfaces to CRM, ticketing, ERP, and other business systems. No manual export is required, and custom data formats and callback endpoints are supported.
Use the wearable reSpeaker Clip — 360° pickup at about 3 m, 14–18 hours of recording without a phone nearby, and audio uploaded through the phone app for transcription. See In-Person Conversation Intelligence for the wearable deployment.
Yes. The Seeed devices used in this solution can be customized with your own logo, enclosure, packaging and pre-loaded firmware, and existing models can be adapted to add or drop interfaces and features. See Customization Service for scope and process, or .