Talk2Nature
NATURE STATION · EARLY PROTOTYPE
Give a spare phone a new purpose. Listen to a place, notice its rhythms, and build a record of the moments you share.
Start with a quiet observation. Mark when you spoke or what you saw. Explore the sounds that happened around it, without assuming they were replies.
Open your station ↓A network starts with one encounter.
A MOMENT OF ATTENTION
Choose a moment. Keep this screen open. Let life happen.
Short clips of louder sounds; quiet periods are not saved.
Describe what you saw. Leave the animal unknown when unsure.
These details stay with this session. Save and discard the session before choosing another animal.
Your observation, not an automatic identification. Use no personal names or exact locations. Example sessions keep these as practice labels.
Nothing uploads or auto-saves. Export before closing to keep your audio and notes.
Phone microphones may filter or change sounds. This app does not identify animals.
On this device · foreground only · playback is always your choice.
The level meter is not calibrated and does not identify animals.
Keep this label general; avoid personal names and exact places. Sensitivity applies to highlights only. Whole moments use no trigger. Neither mode identifies animals.
KEEP A LITTLE CONTEXT
Review the source before sharing. “After my voice” never means “answered me.” Use headphones away from animals for manual playback.
Your observation will appear here after recording.
Save session downloads one ZIP containing your journal and retained WAVs. Check the download before leaving. Nothing is saved automatically. Reopen the ZIP in Compare moments, or unzip to review a WAV in the sound desk. This is not an admitted research dataset.
THE LARGER IDEA
An experimental native Android project now has a user-started foreground service, visible Stop control, 1-, 4- and 8-hour limits, encrypted bounded local storage and manual review/export. It is not a signed APK or field-tested release; physical-device checks are the next gate.
Read the source and build limits ↗Learn recurring sounds, their timing and observed context. Compare what follows a person’s voice with matched quiet periods. Retain absence and uncertainty, not just interesting clips.
A scientist-led Signal Lab could explore a small, voluntary exchange. Generated calls and adaptive playback need controlled response evidence and welfare review before animal use.
The browser prototype runs a simple level detector, and the Android source is a local-only prototype. Uploads, trained encounter models, TV pairing, smart-speaker integrations and animal translation are not implemented.