Where SoundGuessr's sounds come from
SoundGuessr's sounds come from players, who record places and upload them, and from openly licensed collections. Imported field recordings and imported bird, animal, instrument and vehicle clips come from Freesound. The language rounds use Mozilla Common Voice and Google FLEURS. Place names for imported recordings come from GeoNames. Every sound is credited after you guess it.
Which sources does SoundGuessr use?
| Source | License | Used for |
|---|---|---|
| Freesound | CC0 or CC BY | Field recordings for the map; birds, animals, instruments, vehicles and machines |
| Mozilla Common Voice | CC0 | Language rounds |
| Google FLEURS | CC BY 4.0 | Language rounds |
| GeoNames | CC BY 4.0 | Place names |
| Players | Shared by the player who made the recording | Any mode the recording fits |
Why only CC0 and CC BY?
Both licenses allow commercial use. CC0 means no rights reserved: anyone can use the sound. CC BY lets anyone use it as long as the author is credited. Sounds under licenses with NC (non-commercial) or SA (share-alike) terms aren't used. For Freesound, the import tool only searches for CC0 and CC BY sounds and checks the license of each sound again before taking it.
How are Freesound recordings picked and named?
For the map rounds, only Freesound sounds with a geotag are imported. A geotag is a position the uploader added on Freesound. The place name comes from GeoNames' list of towns and cities. It is the biggest place within 15 km of the geotag, or the nearest one within 50 km. A sound with no place within 50 km isn't imported. The import also limits how many sounds it takes from one Freesound user and from one place, so a single recordist or city can't fill the map.
Freesound clips imported for Name the sound, like a bird or a violin, don't keep their position, because those rounds are about what you hear. These clips are tidied up on import: silence at the start is cut, long pauses are shortened and the volume is evened out, so the first second, which is all you hear at first, has the sound in it. Field recordings for the map are left as they are.
Where do the language clips come from?
From two open speech datasets. In Mozilla Common Voice, volunteers record themselves reading sentences and give the recordings away under CC0. SoundGuessr takes at most one clip per speaker. Google FLEURS has read speech in 102 languages under CC BY 4.0. Its clips get the same tidying, so the first second is speech.
Recordings of speech never come up in On the map, Pick from 4 or the daily five, because the language would give the place away.
How are sounds credited?
After each round, the answer panel shows a credit line. What it says depends on where the sound came from:
| Source | Credit line |
|---|---|
| Freesound | The sound's title, the Freesound username, a link to the sound and its license, such as CC BY 4.0 |
| Mozilla Common Voice | Mozilla Common Voice (CC0) |
| Google FLEURS | Google FLEURS (CC BY 4.0) |
| A player's upload | Recorded by, then the player's username |
If a player deletes their account, rounds that were already played on their sounds say "Recorded by a former player" instead.
The credit only appears after you guess, since a title can name the place. The game's credits page lists every source with its license. Clips may be shortened and re-encoded for the game.
How do player uploads fit in?
Players record places and upload the files. Every upload is reviewed before other players hear it, and the server removes metadata such as a phone's GPS position from the file. The steps and the rules are in how to record and upload a sound.
How many sounds are there?
On 5 October 2026 the map had recordings from 156 places in 55 countries. Name the sound had 1,137 language clips with 49 languages to choose from, 219 bird clips, 144 instrument clips, 111 animal clips and 170 clips of vehicles and machines. The home page shows the current numbers. To see how a guess on any of them is scored, read how scoring works.