Skip to content

Downloading Recordings from a Project Backup

A project backup is a ZIP of CSV files. The audio is linked, not bundled: each row of recordings.csv has a url that downloads that recording, and templates.csv and soundscapes.csv (when your project has them) work the same way. This article shows how to download all of those files at once with a small script that needs nothing installed.

Good to know before you start:

  • The links work for about 7 days after the backup is made. After that, request a new backup.
  • It takes time: the script starts one download every 30 seconds (about 2,880 files a day), to keep the load on our servers low. See How much space and time you need.
  • No sign-in is needed. Anyone with the backup can use the links, so keep it private.
  • Files keep the original name they were uploaded with (for example 2MM09808_20260327_080400.flac). Recordings are downloaded as .flac, a lossless audio format.
  • A small number of recordings in some older projects (uploaded before 2023) are stored in a legacy format and can only be downloaded while signed in. The script lists them for you (see Recordings that need a sign-in).

How much space and time you need

Space depends on how long your recordings are and their sample rate. As an example, one-minute recordings at 24–48 kHz are about 1.2–2.4 MB each (around 1.5 MB on average), so 10,000 of them need about 15 GB.

Time: the script starts one download every 30 seconds, whatever the speed of your connection. That is about 2,880 files a day:

Recordings Time
1,000 about 8 hours
5,000 about 42 hours
10,000 about 3.5 days
100,000 about 35 days

Because the links expire after about 7 days, one backup covers about 20,000 files. For a bigger project, let the script run until the links expire, then request a new backup and run the script on it with the same output folder: files you already have are skipped and it continues with the rest.

The script's dry run checks a few of your own files and estimates both the space and the time for you, and warns you if the links will expire before it can finish (see step 3).

1. Download the script

Nothing to install: there is one script for Windows and one for Mac and Linux, and each uses only tools your computer already has.

Open the script's page on GitHub (you don't need a GitHub account) and click the script for your computer:

  • For Windows 10 or 11: arbimon_backup_download.ps1
  • For Mac or Linux: arbimon_backup_download.sh

Then click the Download raw file button (↓) at the top right of the file.

2. Put the script next to your backup

Extract your backup ZIP (Windows: right-click it and choose Extract All; Mac: double-click it). Move the script into the folder that contains the backup's CSV files (for example recordings.0001.csv and sites.0001.csv).

To save the audio on an external drive, move the whole backup folder to that drive first, or use the option to choose the output folder (below).

3. Run the script

For Windows: open the backup folder in File Explorer, click the address bar at the top, type powershell and press Enter. A PowerShell window opens in that folder. First, see what will be downloaded, how much space it needs and how long it will take (this takes about a minute):

    powershell -ExecutionPolicy Bypass -File arbimon_backup_download.ps1 -DryRun

Then start the download:

    powershell -ExecutionPolicy Bypass -File arbimon_backup_download.ps1

(-ExecutionPolicy Bypass lets Windows run a script you downloaded, for this one command only.)

For Mac: open Terminal (search for "Terminal" in Spotlight), type cd (with a space after it), drag the backup folder onto the Terminal window and press Enter. First, see what will be downloaded, how much space it needs and how long it will take (this takes about a minute):

    bash arbimon_backup_download.sh --dry-run

Then start the download:

    bash arbimon_backup_download.sh

For Linux: the same two commands as for Mac, in a terminal in the backup folder.

The script prints how many files it found, when the links expire and how long it expects to take, then a line for each file it downloads. When it finishes it prints a summary.

Useful options

Mac / Linux Windows What it does
--out PATH -Out PATH Where to save the files (default: a downloads folder inside the backup folder). For example --out /Volumes/MyDrive/my-project or -Out E:\my-project.
--backup-dir PATH -BackupDir PATH Use a backup folder somewhere else, instead of the current folder.
--only recordings -Only recordings Download only recordings (or templates, or soundscapes).
--limit 20 -Limit 20 Download only the first 20 files, as a quick test.
--dry-run -DryRun Count the files and estimate the size and time; download nothing.
--interval 60 -Interval 60 Seconds between the start of two downloads. The default, 30, is also the minimum: you can make the script slower, not faster.

What you get

    downloads/
    ├── recordings/
    │   ├── 90326/          one folder per site ID
    │   ├── 90327/
    │   └── sites.csv       which site name each site ID is, and how many files it has
    ├── templates/
    ├── soundscapes/
    └── download-report-….csv   only if some files need attention

Recordings are grouped in one folder per site ID; recordings/sites.csv (and sites.0001.csv in your backup) shows the site name for each ID. Long uploads that Arbimon split into parts have a part number in the name (for example name.003.flac). If two recordings in the same site have the same name, the second one gets its recording ID added (name__313708070.flac), so nothing is overwritten.

Stopping and resuming

Large projects take days, so keep your computer on and awake while the script runs. You can stop it at any time with Ctrl + C and run the same command again later: it skips files that are already complete (without contacting our servers) and continues a half-downloaded file where it stopped. If you restart it straight away, it first waits out the rest of the 30 seconds since its last download. Running it again after it has finished is also a quick way to check that every file is there.

Keep the hidden .arbimon-download-state.tsv file in the downloads folder until you are done; it records what has been downloaded. (If you started with the earlier Python version of the script, the new scripts pick up its progress too.)

If the backup is more than about 7 days old, the script stops with the message "The download links in this backup have expired". Request a new backup on the Project Settings page and run the script on the new backup with the same output folder: files you already have are kept and skipped.

Recordings that need a sign-in

Some recordings in older projects can only be downloaded while signed in to arbimon.org, so the script cannot download them. It counts them in the summary and lists them, with their links, in download-report-….csv. Open each link in your browser while signed in to Arbimon, or contact us if there are many.

Getting help

If the script reports failed files, run it again: it retries only what is missing. If problems continue, email [email protected] with the last lines the script printed and the download-report-….csv file, if there is one.

Interested in pattern matching detections? The CSV you download from a pattern matching job (download icon or Export all results) lists every match with its recording, its start and end time in seconds (x1, x2) and a link to the recording's audio, so you can open or cut just the detected part in any audio editor; see Creating a Pattern Matching Job.