With the right settings and a suitable recording environment, you can significantly improve the detection accuracy of BirdNET Live. Many parameters only need to be set up once and can then be left unchanged.
Tips for better detections
- Hold your smartphone as steady as possible and point the microphone at the sound source.
- Avoid conversations or other loud noises during recording.
- Whenever possible, use Live Mode in a quiet environment with little traffic noise or other background sounds.
- Protect the microphone from wind, as wind noise can impair detection.
- For the best results, we recommend an external microphone with a windshield. This improves recording quality especially for quiet or more distant animal sounds.
Settings
Microphone
Gain
With the microphone gain, you can digitally increase the volume of the input signal. This is helpful when your microphone records very quietly.
However, too much gain also increases the background noise and can degrade detection quality. Therefore, only adjust the value when it is really necessary.
High-pass filter
The high-pass filter suppresses low-frequency sounds such as wind, road traffic, or footfall noise. This can help bird sounds stand out better in noisy environments.
If the filter is set too high, however, low-pitched animal sounds may also be attenuated.
Select microphone
Here you can choose which microphone BirdNET Live should use – for example your smartphone’s built-in microphone or a connected external microphone.
Inference
In this area, you define how the AI model analyzes the recorded sounds.
Window duration
Determines the length of the window that the AI analyzes at once.
Confidence threshold
The confidence threshold defines the level of certainty at which a detection is shown.
Important: a high confidence value is no guarantee of a correct detection. AI can always make mistakes.
Low value: More detections, but the probability of misidentifications increases.
High value: Fewer misidentifications, but fewer animal sounds are detected.
Sensitivity
The sensitivity affects how sensitively the AI model responds to animal sounds.
A higher sensitivity can detect more calls, but at the same time increases the likelihood of misidentifications.
Inference rate
The inference rate determines how often BirdNET Live runs new analyses.
Short intervals capture brief calls better but require more computing power and battery. Longer intervals save energy but may miss short calls.
Announcements
If you like, BirdNET Live can automatically announce newly detected animal species.
For this feature, we recommend using headphones. That way the voice output is not picked up again by the microphone and does not affect the analysis. The announcements are especially helpful when you carry your smartphone in your pocket during a hike.