Everything you need to know about heardhere — how it works, what it detects, and what to expect.
A heardhere unit is a weatherproof, solar-powered acoustic monitoring station. It listens to your site 24 hours a day — identifying birds by their calls at dawn and throughout the day, and detecting bats by their ultrasonic echolocation calls at dusk and through the night. All detections are uploaded in real time to a live website unique to your location, along with a daily AI-written field note summarising what was heard.
No. The unit identifies species from sound, but only the result — a species name, a confidence score, and a timestamp — is ever uploaded or shown anywhere, including to the station's own owner. The recorded audio itself is never uploaded to heardhere, never stored on our servers, and never played back on the website, the owner dashboard, or the on-site kiosk display.
We take this seriously because an outdoor microphone can pick up nearby conversation as well as wildlife. On the device itself, the software briefly holds a short window of raw audio while identifying a species, then discards it automatically — never later than the following day, and typically much sooner. See our Privacy Policy for the full detail.
We don't have a tested figure for this, and we'd rather say that plainly than offer false precision. The unit's microphone hasn't been measured for human-speech pickup range — it's only ever been evaluated for detecting bird and bat calls, which is a different task with a different acoustic profile.
Worth not conflating the two: our bird-detection range (see "What is the detection range?" below) reflects BirdNET's ability to algorithmically pull a loud, distinctive call out of background noise at distance — it isn't a measure of raw audibility. Ordinary conversation is quieter at the source and sounds very different, so it doesn't scale the same way. A mic sensitive enough to catch birdsong at range would plausibly also pick up nearby conversation, at a shorter but still real distance.
Because we can't responsibly give you a "safe beyond this distance" number, our siting advice is the reverse: assume any conversation within earshot of the unit could be picked up, and position it facing away from seating areas, doorways, and other spots where people gather and talk.
No mains power needed — the unit runs on solar with battery backup. It can connect either via 4G (no WiFi needed at all — genuinely remote sites like woodland edges, field margins, and upland areas with just reasonable phone signal work fine) or over your own WiFi network if you'd rather not pay for a data connection.
Either a 4G SIM (data usage is modest — typically a few hundred megabytes per month — and we manage the SIM as part of the platform subscription) or your own WiFi network, if you'd rather avoid the data cost. The unit uploads bird and bat detections every 15 minutes either way.
No. On your WiFi it behaves like any other client device — a smart TV or a phone, say — it joins the network to reach the internet and nothing more. It doesn't scan for other devices, listen to other traffic, or run anything capable of doing so; there's no port scanning, packet sniffing, or network discovery software on it at all. It only ever opens outbound connections to our own servers (and, for the daily AI field note and the optional camera's human/pet filter, to Anthropic's API) — it never accepts inbound connections from your network, so nothing on your side can reach it either, and it can't be used as a way in to anything else you have connected.
If you'd rather it not touch your network at all, the 4G option sidesteps the question entirely — it uses its own separate mobile data connection and never joins your WiFi.
The unit is housed in a weatherproof enclosure and arrives fully built — nothing to assemble or wire up. You mount it yourself on a post or tree at approximately head height, with the solar panel positioned to catch maximum sun, then power it on and scan the QR code on the label to configure it. The whole process takes about five minutes and needs no specialist equipment or a visit from us.
Battery life depends on your local weather and time of year — the solar panel is sized to keep the battery topped up in typical UK conditions, but several consecutive overcast days, especially in deep winter, can draw it down. If you notice your station going offline overnight, let us know and we'll take a look.
Bird calls are identified using BirdNET, a deep learning model developed by the Cornell Lab of Ornithology and Chemnitz University of Technology. BirdNET analyses 15-second audio clips and matches them against a library of over 6,000 species. Each detection includes a confidence score — we typically display detections above 70% confidence.
BirdNET performs well on common UK species but like all AI models, it can occasionally misidentify species with similar calls, particularly in noisy conditions. Treat the data as strong evidence rather than certainty.
Bat echolocation calls are recorded by an AudioMoth ultrasonic microphone at 384kHz — well above the range of human hearing. They are then identified using BatDetect2, a neural network trained on UK bat species. Detection confidence scores are shown on the bats page.
Bat identification by call is inherently more challenging than bird identification — some species have overlapping echolocation frequencies, and call characteristics vary with behaviour. Results are best interpreted at the species group level (e.g. Pipistrelle species) rather than always at exact species level.
Yes — owl calls fall within the frequency range detected by the bird microphone and BirdNET includes Tawny Owl, Barn Owl, Little Owl and other UK species. Bird detection runs until sunset, when the unit switches over to bat detection for the night, so owl calls from earlier in the evening and pre-dawn are captured normally.
Effective detection range varies by species and conditions. Loud, clear singers like Chaffinch or Blackbird may be detected at 50–100 metres in still conditions. Quieter species or calls in wind or rain will have a shorter effective range. For bats, the AudioMoth at 384kHz typically detects echolocation calls within 10–30 metres depending on species and call intensity.
Yes — wind noise and heavy rain reduce detection sensitivity and can generate false positives. The AI models are generally robust to mild background noise but performance drops in severe weather. This is reflected in the detection data — you'll notice fewer detections on stormy days, which is ecologically meaningful as well as technical.
Detection data uploads every 15 minutes. The live station page refreshes automatically, so you'll see new detections appear without reloading. The AI Naturalist field note is generated once per day at approximately 08:00, covering the previous 24 hours of acoustic activity.
Each morning, an AI (Claude, by Anthropic) reads the previous 24 hours of bird and bat detections and writes a short field note in the style of a nature journal — warm, precise, and ecologically informed. It's not a summary of statistics; it's an attempt to interpret what the acoustic record means for the life of the site.
The AI Naturalist draws on the detection data only — it doesn't have eyes or ears at the site. Think of it as a thoughtful observer making sense of the numbers, not a replacement for human expertise.
Yes — each station has a public URL (e.g. dam-mire.heardhere.uk) that anyone can visit. You can share it freely, embed it in your own website, or link to it from social media. We also provide embed codes for displaying a detection count, species badge, or naturalist note on external websites.
All detection data is stored and the activity calendar on the birds and bats pages shows the full history since the unit was installed. Owners can request a CSV export of raw detection data via the owner dashboard. Birds and bats pages also show phenology tracking — first-of-season arrivals and year-on-year species comparisons — once enough seasons of data have built up.
The bird microphone (AB13X) and the bat microphone (AudioMoth at 384kHz) share the same USB bus on the Raspberry Pi that powers the unit. The AudioMoth's ultrasonic recording at 384kHz requires significant USB bandwidth — enough that both microphones cannot operate simultaneously.
Bird detection runs from an hour before dawn until sunset, then hands over to bat detection through the night — the switch itself takes only seconds, not an extended gap.
The unit is designed to recover automatically from power loss. On restart, it checks what time it is, determines whether it should be in bird or bat detection mode, and resumes immediately. Detections made while offline are stored locally and uploaded in batches once connectivity is restored. The station page will show the last known update time if the unit has been offline.
Our team can reach a unit directly for remote support and diagnostics if you ever run into an issue — a private, outbound-only connection to that device alone, never a route into anything else on your network. Day to day, you don't need direct access to the unit itself — your owner dashboard covers everything you need to manage your station.
heardhere is built on the shoulders of excellent open-source projects. Bird identification uses BirdNET-Pi (Cornell Lab of Ornithology / Chemnitz University). Bat detection uses BatDetect2 via the acoupi framework. Audio recording uses AudioMoth from Open Acoustic Devices. We're grateful to the teams behind all of these.
Get in touch via the contact form. Tell us about your site — where it is, what habitat, and what you're hoping to monitor. We'll come back to you within a couple of days to discuss the best setup for your needs and talk through installation.
Absolutely. The live station website is designed to be accessible to anyone — no technical knowledge needed. Schools can use the live data to support nature study and citizen science projects. The AI Naturalist field note gives daily context that makes the data meaningful for non-specialists. We're happy to discuss educational partnerships and reduced-cost arrangements for schools and community groups.
heardhere is built on open-source components and we're a small team who believe in openness. If you're technically minded and want to self-host, get in touch — we're happy to point you in the right direction. The managed service (unit + platform + support) is designed for those who want the monitoring without the technical overhead.
All UK bat species are legally protected, and evidence of bat activity is often required before development, felling, barn conversion, or other building work near roosting or foraging habitat. Continuous monitoring gives you a genuine, dated record of activity — real evidence, not a one-off survey snapshot. We're not a substitute for a licensed ecologist's formal survey, but the data is a valuable, low-cost complement to one.
Yes — continuous species and activity data over time is exactly the kind of evidence funders and stewardship schemes (Countryside Stewardship, ELMS, and similar) look for. Instead of a single ecologist visit, you get a dated, ongoing record showing real trends — proof that habitat management is actually working, not just a claim that it is.
Yes — continuous, dated monitoring data is a genuinely useful complement to Biodiversity Net Gain evidence on a development site, showing real activity over the project period rather than a single site visit. Since most developments only need coverage for a defined period rather than permanently, seasonal rental is often the better fit — see our pricing for details.