Talk2NatureBioacoustics & behaviour
Which observations, context and uncertainty does your team need to keep attached to a sound?
FOR RESEARCH GROUPS
One small test.
One honest answer.
Compare two invented sessions in your browser. See what the records support. The Audacity handoff is optional.
A few minutes to start · optional ~20-minute handoff · no account to try · no animal recording · no upload
Finished a review? Share a short note ↗. One sentence is enough; device details are optional. Posting requires GitHub sign-in, and reports are public. For a detailed software rerun, use the workflow report.
WHY ASK A UNIVERSITY?
Research teams know
where the seams are.
The best contribution may be a short review, a failed run or a clear decision that an existing tool is already better. We need the workflow reality before building more software.
Research software & libraries
Can another person reproduce the instructions, inspect provenance and identify information loss?
Accessibility & field methods
What fails on the devices and in the places where researchers actually work?
READY FOR REVIEW
Invented audio.
Real workflow feedback.
- Local browser workflow and synthetic test sound
- Timed labels, original notes and conversion report
- Exact revision and software version can be recorded
- One small task; negative results are useful
NOT READY FOR A FIELD STUDY
First, earn the
scientific gates.
- No Talk2Nature animal-communication model
- No accepted research recordings or public data intake
- Ultrasonic and synchronized audio/video workflows are unverified
- No independent scientific reviewer or institutional protocol approval
A USEFUL FIRST CONTRIBUTION
Show us one thing
the tool gets wrong.
Try the example, then send a short, redacted report: what you tried, the device and software versions, what you expected, what happened and whether the workflow solves a real task. Do not send recordings or private research data.
Independent evaluation is welcome; endorsement, partnership and study data are not requested. We will not name or quote your group without permission.
Optional: review a scientific claimThree bounded research and measurement questions
OPEN METHODS REVIEWS
Check one claim
against the evidence.
The browser exercise tests software; it does not validate the biology. Researchers can review receiver-response evidence, limits of phone recording as measurement, or the honeybee signal-and-follower candidate.
THE BOUNDED QUESTION
What do the cited receiver-response experiments support about call sequence and order, and which competing explanations remain unresolved?
- Check one statement against the cited paper and exact section, table or figure.
- Flag an alternative explanation, missing control or limit on generalization.
- Name the smallest public-evidence check that could change your assessment.
This is a literature critique, not a request to plan or conduct playback, share recordings or unpublished data, endorse the project, or join a partnership. Reports are public; do not include personal details or sensitive locations.
THE BOUNDED QUESTION · MOBILE AUDIO
Could a small no-animal reference test tell us anything useful about phone and app-path variability without being mistaken for calibration?
- Separate microphone, operating-system/browser processing and app-path effects.
- Decide whether reference recordings and task-specific uncertainty are useful at all.
- Keep device metadata minimal and avoid animal, participant or location data.
This is a methods critique, not an approved protocol or request to collect wildlife audio. Physical Android/iOS behavior, including iOS Safari, remains unverified.
THE BOUNDED QUESTION · HONEYBEE SIGNAL AND RESPONSE
Could a bee-held-out analysis add useful evidence beyond the published links between waggle-dance features, follower behavior and feeder outcomes?
- Check which parts of the signal-to-follower-to-outcome chain are directly observed and which are inferred.
- Assess whether bee-held-out evaluation is meaningful and feasible in a two-season, single-site study.
- Identify the strongest prior-art or location-privacy issue that should change the decision.
This is a no-data literature and methods critique. The study's published associations are prior art; one site cannot establish colony- or site-level generalization. No archive, code, model or bee observation is requested, and this is not evidence of general animal translation.
Optional: explore the recording-network ideaWhat is available now and what a shared system would require
THE NETWORK QUESTION
Build a useful network
without collecting everything.
For birds, established tools already support sound identification and large-scale analysis, recording archives, and biodiversity media metadata. That is one mature corner of the field, not a solution for every animal or a way to establish what a signal means. Talk2Nature should connect careful observation context and provenance to existing tools, not ask every lab to upload its audio into a new central bucket. BirdNET-Analyzer · Macaulay Library audio guidance · GBIF download formats.
Review the workflow
Try synthetic sessions, local capture and file tools. Recordings stay with the user; there is no Talk2Nature upload or shared recording server.
Design one study path
First agree on one research question, device path, labels, rights, retention and withdrawal with the institution. Keep any real study files in its approved storage.
Share safe discovery metadata
If several groups need the same thing, test a federated catalog: each institution controls its files; only reviewed, permissioned records become discoverable.
That last design is a proposal, not a service. Talk2Nature's private server is a local, owner-only synthetic-metadata rehearsal. No cloud-hosted research workspace, recording intake or university network is available today.