Optimize AudioClip Settings

How to optimize the settings of your AudioClips to use less RAM and CPU

Here are some settings you’d want for most games get lower RAM usage, disk read and CPU usage. Read more about the other settings here: Unity Manual - AudioClip.

AudioClip - Overall Settings

Compression Format: Vorbis Quality: 60. Which is ~192 kbit/s for Vorbis. You can go lower if you’d want to save space but it comes at a loss of quality. Sample Rate Setting: Should be the same as the sample rate of the platform, which today mostly is 48kHz (this is so that it doesnt have to do a sample rate conversion) when playing. Also, if the sounds are created in 48kHz there is more high frequency content which makes it sound better when pitching down vs 44.1kHz.

❕ Tip: If you select multiple AudioClips, you can see their total “Imported Size” which is the size they’ll have after all your format and quality settings are applied.

AudioClip - Vorbis Quality

Unity Vorbis Quality 1 is VBR Quality 0 (not -2). Mono is half the bitrate of stereo and 4 channel files (for surround) are twice the bitrate of stereo.

Unity Vorbis Quality	VBR Target kbit/s stereo	VBR Target kbit/s mono	VBR Range kbit/s stereo	Noise Test kbit/s stereo	Size vs 10	Size vs 60	Comment
100	500	250	500 - 1000	554	100%	260%	Uneccessarely big, don’t use this
90	320	160	320 - 500	435	64%	167%	Uneccessarely big
80	256	128	256 - 320	382	51%	133%	Use if you want to mazimize quality & have the disk space and RAM
70	224	112	224 - 256	309	45%	117%
60	192	96	192 - 224	232	38%	100%	Use this as a default, it is a good balance between quality and size
50	160	80	160 - 192	173	32%	83%	Sounds like near CD quality
40	128	64	128 - 160	128	26%	67%
30	112	56	112 - 128	117	22%	58%	Smaller & better sound than .mp3 128kbps
20	96	48	96 - 112	102	19%	50%
10	80	40	80 - 96	87	16%	42%	Sounds acceptable. Use if you are desparate for saving space and RAM
1	64	32	64 - 80	71	13%	33%	Looses a lot of quality, noticeably stereo separation and bass

Notes: Noise test is calculated using VBR export & size comparison to CBR with a known bitrate. Sources: Recommended Encoder Settings and Vorbis Quality. The bitrates are approximate.

AudioClip - Load Type

Stereo files are double the size of mono files. So a 60s 2 channel sound is the same size as a 120s 1 channel sound. The lengths are approximated with stereo files in mind.

Length Seconds ~	Play Frequency	Load Type	Pros	Cons	Example Sounds
Short - Less than 1s	High	Decompress On Load	Uses less CPU	Uses 10x RAM so only use for short sounds.	Player sounds, footsteps, impacts etc. Commonly used for mobile platforms
Short - Less than 1s	Low	Compressed In Memory	Uses less RAM
Medium - More than 1s less than 60s	High & Low	Compressed In Memory	Uses less RAM
Long - More than 60s	High	Compressed In Memory	Uses less disk read	Try if you have problems with audio delays. Watch out for load times and RAM usage	Localized 3D ambiences etc
Long - More than 60s	Low	Streaming	Uses less RAM	Many platforms can only handle less than 10 simultaneously streaming sounds	Music, background 2D ambiences etc

Note: You want to stream the largest files you have, but not too many at the same time. So the 60s threshold for long sounds can be lowered to e.g. 30s or even lower if not many sounds are +60s.

AudioClip - Preload Audio Data Makes the sound load at startup, which might create long load times on harddrives and slower systems. Is useful to combine with Load In Background so it doesn’t interrupt the main thread.

Length in seconds	Play Frequency	Preload Audio Data	Pros	Cons	Example Sounds
Short - Less than 1s	High	Yes	Less frequent disk read	Use only for frequent & short sounds. Combine with Load in Background	Player sounds, footsteps, impacts etc
Short - Less than 1s	Low	No
Medium & Long - More than 1s	Low	No

Note: I’ve had problems where if I don’t Preload Audio Data, there might be clicks heard in the end of sounds. "Prevent End Clicks" in the SoundContainer mitigates this.

AudioClip - Load In Background When enabled, the AudioClip is loaded without stalling the main thread and any playing is delayed until the AudioClip is loaded. Useful to combine with Preload Audio Data so it doesn’t interrupt the main thread.