WAV 16-bit PCM vs 32-bit float: what audio detach returns

Sume audio detach returns wav as pcm_s16le, which is 16-bit PCM, by default. The docs list no 32-bit float output option, so footage recorded as 32-bit float comes out as 16-bit. Adobe's Premiere 26.5 notes mention support for Sony FX5 cameras including 32-bit float Hi-Res audio; that is what the camera and Premiere handle, not what Sume returns.

Sume facts are from Audio detach, read 2026-09-30.

What is the difference between 16-bit PCM and 32-bit float?

16-bit PCM stores each sample as a 16-bit integer, with a fixed range. 32-bit float stores each sample as a floating-point number, which keeps far more headroom, so a clip that was too loud at capture can often be turned down later without the clipping you would get from integers. Both are plain wav; only the sample format differs. This is general audio background, not a Sume claim.

What can I control on Sume?

When is 16-bit enough?

The docs describe the default wav as what timeline_create audio.url, POST /v1/timeline-1.0/audio and speech-to-text want. For transcription, captions, or a spine for a Timeline render, 16-bit is the supported shape. If you need to regrade levels in post, keep the camera original and do that work from it, not from a detached copy.

What are the limits?

The source can be up to 1800 seconds and the output up to 900 seconds, so a longer track needs a range. A source with no audio track fails with detach_source_has_no_audio. The source video must be a media.sume.com artifact, so import it first. For the sample-rate and channel choices in more depth, see audio detach: keep source channels and sample rate.