Six steps, start to finish
The app walks you through the same sequence every time it opens, so there's nothing to remember between complaints.
Install required components
On first run, the app checks for Python, ffmpeg, and the Python packages the analysis needs, and downloads anything missing directly from python.org and GitHub. Nothing routes through winget or an app store, so it works on locked-down council machines. A live log shows exactly what's happening at each step.
Choose your recordings folder
Point it at the folder holding your WAV files. Tick a box to include subfolders, and the app counts the files it found before you go any further — so you know it's looking at the right recordings before anything runs.
Pick where results are saved
Save straight into a results folder alongside the recordings, or choose a separate export location — useful when recordings live on a shared drive but reports need to go somewhere else entirely.
Set the detection sensitivity
Five presets from Very Sensitive to Very Strict, each with a plain-English description of the trade-off. Outdoor or long-range recordings usually call for a lower threshold; close-range, clear recordings can run stricter with fewer false positives.
Review and run
A summary confirms the folder, sensitivity, and file count, alongside an estimated run time before you commit. Then YAMNet — Google's audio classification model — works through every file, with a live progress bar and log so you can see it moving.
Open the finished report
A formatted Excel workbook, one MP3 clip per detected bark, and a text file of seven ready-to-use prompts — for a full report, a formal notice, a plain-English summary, or a case-file entry. One click opens the results folder. Worth a listen through the flagged clips before anything goes into a formal notice — the AI does the listening, but a person should always do the final check.
Curious what the screens actually look like?
A look at each step of the wizard in the app itself.