Short answer: very good — and getting better.

What powers the IDs

Terra Station uses the BirdNET 6K model, a proven, globally used bird identification model, to identify birds by their songs and calls. If you've used other BirdNET-based tools, Terra's identifications are at least equivalent to anything else built on BirdNET — and the Terra Station's dedicated outdoor microphones are listening for you 24/7, whether you're streaming or not.

Here's the basic process:

1. The Terra Station samples audio continuously from both of its microphones.
2. The audio is analyzed by the BirdNET 6K model.
3. When a bird is detected, the sound snippet is saved with the species, confidence, and date and time, and shows up in your app and on the website.

What to expect

No sound-ID system is perfect, and occasional misses or misidentifications are normal — a distant, faint, or partly obscured song is hard for any listener, human or machine. Two things make a big difference:

• Placement. What matters most is how loud the birds are compared to the background noise. A station near cover, feeders, or water — with road noise blocked by the house or a hill — will identify more, and more accurately, than one out in the open near a road. See Where are some good places to put my Terra?
• Location (GPS). Your Terra Station uses its GPS position to apply the right regional expectations for your part of the world. It needs at least a partial view of the sky when first placed to set its coordinates. If ALL your IDs look wrong for your region, that's usually a GPS issue — get in touch at support@terra.zohodesk.com and we'll check it.

Actively improving

Bird ID on Terra is not standing still. We're actively improving the identification engine through customer feedback and ongoing engine development, including work on a separate model with better accuracy — and other models will be available in the future. We're also working on modifications that will allow frog and insect detections.

Spotted an ID that seems off, or a bird you heard that Terra missed? We'd genuinely like to know — send us a message at support@terra.zohodesk.com with the date, time, and species, and it helps us make the engine better for everyone.