PCM audio is the largest format on disk and in memory

Compression Format: PCM stores the clip as raw samples — the largest option both in the build and in memory. It exists for the narrow case where zero decode latency matters more than size. On mobile it is almost never the right choice, and it’s often set by accident when someone was reaching for a Load Type setting instead.

What the scan looks at

Audio clips whose Compression Format is PCM, on the active build target’s settings.

Why it costs you

Load Type decides when decoding happens; Compression Format decides how big the data is. PCM makes the compression question moot — the data is already raw, so there’s nothing to decode and nothing saved. A stereo minute of 44.1 kHz PCM is around 10 MB in the build; the same clip as Vorbis is a fraction of that with no audible difference at normal quality settings.

The formats worth knowing:

• Vorbis — good general default for music and SFX. Quality slider controls size.
• ADPCM — fixed ~3.5:1 compression, extremely cheap to decode. Good for short, frequent sounds (footsteps, impacts) where Vorbis decode per voice adds up.
• PCM — only for very short sounds that must start with zero decode work.

How to fix it by hand

Set Compression Format to Vorbis in the importer (per platform if you’re overriding), reimport, and listen. If a short SFX is retriggered dozens of times per second and the decode shows up in a profile, ADPCM is the compromise — still a fraction of PCM’s size, decode cost close to nothing.

What PerfLint does about it

The finding names the clip, its length, and the current format. Pro can switch it to Vorbis from the row, one clip or the whole group — importer setting, reversible. The scan doesn’t pick ADPCM for you, because “this sound is retriggered constantly” isn’t something a static scan can see.