---
canonical: https://biosonic.io/compare/tadarida
---

# BioSonic vs Tadarida and Vigie-Chiro: open-source and commercial bat identification compared

Last updated: 2026-11-18. Written by the BioSonic team.

Tadarida is a free, open-source toolbox (Bas, Bas and Julien, 2017) that detects, labels and classifies animal sounds, and it produces the bat identifications for Vigie-Chiro, the French national citizen-science bat monitoring programme run by the Muséum national d'Histoire naturelle. BioSonic is a Swedish browser platform that identifies European bat species from WAV, FLAC or Anabat recordings with a ready-trained model, so there is nothing to install or train. Choose Tadarida or Vigie-Chiro for an open, transparent research tool and the French national protocol; choose BioSonic if your team needs results and shared review without building its own classifier.

There is no published benchmark that compares BioSonic and Tadarida on the same recordings, so this page compares what each is built for, what it takes to use, and how the two can work side by side.

## At a glance

Facts about Tadarida come from the authors' paper and GitHub repositories; facts about Vigie-Chiro come from vigienature.fr. All are listed under [References](#references).

| Feature | BioSonic | Tadarida / Vigie-Chiro |
|---|---|---|
| What it is | Browser platform for automatic bat species identification and team review | Tadarida: open-source toolbox for detecting, labelling and classifying sounds [1]. Vigie-Chiro: French citizen-science bat monitoring programme that runs Tadarida on uploaded recordings [5][6] |
| Made by | Biosonic AB, Stockholm, Sweden | Tadarida: Yves Bas, Didier Bas and Jean-François Julien [1]. Vigie-Chiro: MNHN and OFB, part of Vigie-Nature, since 2006 [5] |
| Cost and licence | Commercial service; free trial, then a package of analysis gigabytes | Free and open source: Tadarida-D and -L under LGPL-3.0, Tadarida-C under GPL-3.0 [1][2][3][4] |
| Set-up | Runs in the browser; nothing to install | Command-line detector (C++/Qt), labelling interface, and R scripts with randomForest and data.table [1][4]; Vigie-Chiro participants use a web portal [6] |
| Ready-trained classifier | Yes, trained on 2.5 million verified bat calls and 3.5 million noise files; new model every weekend | Inside Vigie-Chiro, yes. On your own, you train it on a labelled reference database; the public sample classifier is described by its author as poor [4] |
| Method | Deep learning image recognition on spectrograms | 269 features per sound event, classified with random forests [1] |
| Species and region | European bat species | First designed for French common bats and bush-crickets [1]; Vigie-Chiro covers metropolitan France [6] |
| Output | Species suggestion and confidence score per call, up to 7 species per recording, reviewed on the spectrogram | Species per file with confidence values (.tc files) [1][4]; Vigie-Chiro gives a confidence index from 0 to 1 [6] |
| File formats | WAV and FLAC from full-spectrum detectors, any file length; Anabat zero-crossing converted on upload | WAV recordings [1]; Vigie-Chiro asks for recorder settings set by the MNHN [6] |
| Team review | Whole team in shared projects, every confirmation or correction attributed | Vigie-Chiro: per-participation results page and validation tutorials; ChiroSurf for detailed checking [6][7] |
| Where data lives | Stored in the EU, mainly in Sweden | Tadarida: on your own computer. Vigie-Chiro: uploaded to the programme's web portal [6] |
| Transparency | Method and benchmark results published on our benchmarks page | Full source code and method paper are public [1][2][3][4] |
| Support | Email support from the BioSonic team | "Support is not guaranteed", though users are encouraged to contact the authors [1] |
| Published accuracy | 98.7% F1 in a third-party benchmark by Wilder Sensing | No overall bat accuracy figure in the paper; error rates depend on your reference database [1][4] |

No benchmark has tested BioSonic and Tadarida on the same recordings. The two accuracy rows above are not comparable with each other.

## How each one works

### Tadarida: an open toolbox in three parts

Tadarida was published in 2017 in the Journal of Open Research Software by Yves Bas, Didier Bas and Jean-François Julien [1]. It has three modules. **Tadarida-D** is a command-line program that finds sound events in WAV files and extracts 269 features from each, saved as .ta files [1][2]. **Tadarida-L** is a graphical interface for labelling those events to build a reference sound database [1][3]. **Tadarida-C** is a set of R scripts that train random forest classifiers on that database and apply them to new recordings, giving species and confidence values per file [1][4].

The toolbox was "first designed for the specific need of the monitoring of French common bats and bush-crickets" [1]. At publication the authors' reference database held 863,730 labelled sound events of 91 species from 17,773 WAV files [1]. Because the code and method are open, anyone can inspect, change or retrain it.

### Vigie-Chiro: Tadarida at national scale

Vigie-Chiro was launched by the MNHN in 2006 to follow bat populations in France and is now run with the OFB as part of Vigie-Nature [5]. The programme now favours its fixed-point protocol, introduced in 2014: a participant records a full night at the same place twice each summer, once between mid-June and the end of July and once between mid-August and late September, and repeats it in later years [5][6]. Files go to the Vigie-Chiro web portal, Tadarida identifies the species, and the participant receives an email with a link to the results [6].

The scale is large. The Tadarida paper reports automated use on Vigie-Chiro data since October 2015, by which time 296 participants had uploaded 12,952,016 WAV files to the portal [1]. Vigie-Chiro is open about the limits: its protocol page says the automatic identifications should be treated with caution because they contain some error and uncertainty, and it offers tutorials and the free ChiroSurf tool for validating results [6][7].

### BioSonic: a trained model and a shared review screen

BioSonic converts each recording into a spectrogram and runs deep-learning image recognition on it. The model is trained on more than 2.5 million verified bat calls and 3.5 million noise files from across Europe, curated by 12 bat experts, and a new model goes live every weekend.

You upload WAV, FLAC or Anabat files of any length in the browser. Noise is filtered out, every call gets a species suggestion and confidence score, and your team clicks through the spectrograms to confirm or correct. Every change is recorded against the person who made it, and detection tables, charts and maps come out of the same project.

## Accuracy and evidence

BioSonic scored **98.7% F1 in a third-party benchmark by Wilder Sensing**, against 83.0% for Kaleidoscope Pro on the same recordings, with 13.4× fewer false positives and 16.3× fewer missed bats. The method and results are on our [benchmarks page](https://biosonic.io/benchmarks) [8].

**Tadarida was not part of that benchmark, and we know of no published test of BioSonic against Tadarida.** We therefore make no claim about which is more accurate. The Tadarida paper does not give an overall accuracy figure for bats; its authors describe a trade-off between error rates for common and rare species, and the Tadarida-C documentation warns that error rates computed on your own reference database could be underestimates of ground truth [1][4]. A classifier's real performance also depends on the reference calls it was trained on, so Vigie-Chiro's classifier and one you train yourself will not behave the same.

Both approaches need a person in the loop. Vigie-Chiro asks participants to treat automatic results with caution [6]. BioSonic sends hard groups, such as Myotis to species level and Plecotus, to manual verification, in line with the Bat Conservation Trust's expectation that automated identifications are checked by a person [9]. The fairest test is to run the same nights through both and compare them with identifications you have verified yourself.

## When to choose which

Tadarida and Vigie-Chiro are strong, open tools built for national monitoring and research. BioSonic is built for consultancies and teams who need finished, reviewed results for projects. Many people will fit one side clearly.

### Choose BioSonic if you

- Want automatic identification without installing software, writing R, or training a classifier
- Run surveys for clients across Europe and need detection tables, charts and maps for reports
- Work as a team and want shared projects with an audit trail of who confirmed what
- Record with different detectors, including FLAC or Anabat zero-crossing files
- Want a model that is retrained every weekend and support from the people who build it

### Choose Tadarida or Vigie-Chiro if you

- Take part in French national bat monitoring and follow the Vigie-Chiro protocol
- Need free, open-source software whose code and method you can inspect and cite
- Are a researcher who wants to train and test your own classifier on your own reference calls
- Work with bush-crickets as well as bats
- Are comfortable with the command line and R, and need no vendor support

### Using both

The two do not exclude each other. Both work on WAV recordings, so a French participant can keep sending fixed-point nights to Vigie-Chiro, where the data feed national trends, and also upload the same files, or separate project recordings, to BioSonic for team review and client reports. Researchers can use Tadarida's open features and classifiers alongside BioSonic's results to see where the two agree and where a human should look closer.

Tadarida also reaches beyond France: the BTO Acoustic Pipeline's bat classifier uses it (see our [BTO Acoustic Pipeline comparison](https://biosonic.io/compare/bto-acoustic-pipeline)).

## Cost

**Tadarida** is free to download, use and change under its open-source licences [2][3][4]. What it costs is time: setting up the tools, labelling reference calls, and the computing power to train classifiers (the paper notes about 15 GB of memory for the authors' full database) [1].

**BioSonic** starts with a free trial. After that, you buy a package of analysis gigabytes; the bigger the package, the lower the price per GB, and recordings are kept for 9 years. There are no per-seat fees. [Ask for a quote](https://cal.com/biosonic/30min).

## Frequently asked questions

### Is Tadarida free?

Yes. Tadarida is free, open-source software with its code on GitHub. Tadarida-D (detection) and Tadarida-L (labelling) are released under the LGPL-3.0 licence and Tadarida-C (classification) under GPL-3.0 [1][2][3][4]. The cost is time and expertise: you set up the command-line tool and R scripts yourself and, outside Vigie-Chiro, build a labelled reference database to train the classifier. The authors state that support is not guaranteed, though users are encouraged to contact them [1].

### What is an alternative to Tadarida for consultants?

If you want automatic bat identification without installing or training anything, BioSonic is one alternative: you upload WAV, FLAC or Anabat files in a web browser, a ready-trained model identifies European bat species, and your team reviews the results together. It scored 98.7% F1 in a third-party benchmark by Wilder Sensing, but it has not been benchmarked against Tadarida. Other commercial options include Kaleidoscope Pro, SonoBat, BatExplorer and SonoChiro; our [list of Kaleidoscope Pro alternatives](https://biosonic.io/kaleidoscope-pro-alternatives) compares them.

### Has BioSonic been benchmarked against Tadarida?

No. We know of no published test that runs BioSonic and Tadarida on the same recordings, so we make no accuracy claim between them. BioSonic's published figure is 98.7% F1 in a third-party benchmark by Wilder Sensing, against 83.0% for Kaleidoscope Pro in the same test ([benchmarks](https://biosonic.io/benchmarks)). The Tadarida paper does not report an overall accuracy figure for bats [1]. The fairest check is to run the same nights through both and compare the results with identifications you have verified yourself.

### How does Vigie-Chiro use Tadarida?

Vigie-Chiro is the French citizen-science bat monitoring programme, launched by the Muséum national d'Histoire naturelle (MNHN) in 2006 and now run with the Office français de la biodiversité (OFB) as part of Vigie-Nature [5]. Participants record a full night at a fixed point, upload their files to the Vigie-Chiro web portal and receive an email when processing is finished. Tadarida provides the species identifications, each with a confidence index from 0 to 1, and the programme asks participants to treat them with caution because they contain some error and uncertainty [6].

### What do I need to run Tadarida myself?

Tadarida-D is a command-line program written in C++ with Qt, tested on Windows and Linux; it detects sound events and extracts 269 features from each into .ta files. Tadarida-L is a graphical labelling tool, tested on Windows, for building a reference sound database. Tadarida-C is a set of R scripts that use the randomForest and data.table packages to train a random forest classifier and apply it [1]. The public repository includes only a small sample reference database, and its author warns that a classifier built from it performs poorly, so you need your own labelled calls [4]. The paper notes that training on the authors' full database used about 15 GB of memory [1].

### Can I use Vigie-Chiro or Tadarida and BioSonic together?

Yes. Both work on WAV recordings, so you can keep taking part in Vigie-Chiro with the protocol it asks for and also upload the same files to BioSonic for project work, team review and reports. Running the same nights through both is also the simplest way to see where they agree.

Prefer another language? [Analyse acoustique des chauves-souris](https://biosonic.io/fr/analyse-acoustique-chauves-souris) explains BioSonic for consultancies in France, Belgium and Switzerland, in French.

## Other comparisons

- [BioSonic vs Kaleidoscope Pro](https://biosonic.io/compare/kaleidoscope): cloud AI vs Wildlife Acoustics' desktop classifier, compared on the same benchmark.
- [BioSonic vs BTO Acoustic Pipeline](https://biosonic.io/compare/bto-acoustic-pipeline): BTO's pipeline, whose bat classifier uses Tadarida.
- [BioSonic vs SonoBat UK](https://biosonic.io/compare/sonobat): cloud review for teams vs a long-standing desktop vetting tool.
- [See all comparisons](https://biosonic.io/compare)

## References

1. Bas, Y., Bas, D. and Julien, J.-F. (2017). Tadarida: A Toolbox for Animal Detection on Acoustic Recordings. Journal of Open Research Software 5(1): 6. [doi.org/10.5334/jors.154](https://doi.org/10.5334/jors.154)
2. Tadarida-D source code (detection and feature extraction, LGPL-3.0): [github.com/YvesBas/Tadarida-D](https://github.com/YvesBas/Tadarida-D)
3. Tadarida-L source code (labelling interface, LGPL-3.0): [github.com/YvesBas/Tadarida-L](https://github.com/YvesBas/Tadarida-L)
4. Tadarida-C source code and documentation (classification in R, GPL-3.0): [github.com/YvesBas/Tadarida-C](https://github.com/YvesBas/Tadarida-C)
5. Vigie-Nature, Vigie-Chiro programme page (history, partners, protocols): [vigienature.fr/fr/vigie-chiro](https://www.vigienature.fr/fr/vigie-chiro)
6. Vigie-Nature, Vigie-Chiro protocol in detail (fixed-point protocol, web portal, Tadarida results and confidence index): [vigienature.fr/fr/le-protocole-en-detail](https://www.vigienature.fr/fr/le-protocole-en-detail)
7. Vigie-Nature, Vigie-Chiro additional resources (validation tutorials, ChiroSurf): [vigienature.fr/fr/ressources-complementaires](https://www.vigienature.fr/fr/ressources-complementaires)
8. Wilder Sensing, third-party benchmark, 5,325 predictions: [biosonic.io/benchmarks](https://biosonic.io/benchmarks)
9. Bat Conservation Trust, guidelines for passive acoustic surveys of bats in woodland (manual verification of automated identifications): [bats.org.uk](https://www.bats.org.uk/our-work/science-research/passive-acoustic-surveys/guidelines-for-passive-acoustic-surveys-of-bats-in-woodland)

Tadarida and Vigie-Chiro facts last checked: 2026-11-18.

## Ready to try it?

Upload a night you have already run through Tadarida or Vigie-Chiro and compare. No commitment, no installation required.

[Start free trial](https://bat.biosonic.se/signup) | [Book a demo](https://cal.com/biosonic/30min)

Questions? Email [josef.carlson@biosonic.se](mailto:josef.carlson@biosonic.se)
