Mux audio-only M4A vs Sume audio detach: which file to send to STT

Mux can produce an audio-only static rendition of a video, delivered as an M4A file. Sume's audio detach does the same job for a video on media.sume.com but returns a wav, which is sample-exact and the default, or an mp3 at 128 kbps. If the audio will be joined again or sent to speech-to-text, wav with 16000 Hz mono is the documented shape.

Mux's side is from its static MP4 renditions guide, read 2026-10-03. Sume's is from the audio detach and timeline audio docs.

What each service returns

The two are built for different jobs. Mux's rendition list is about downloadable MP4 versions of a stream, and its audio-only option is one entry in that list. Sume's detach is a separate step that mints a reusable audio file.

Why Sume defaults to wav

Sume's detach docs call the default pcm_s16le wav sample-exact, and say it is what timeline audio, the timeline audio.url field and speech-to-text want. The timeline audio page adds that mp3 is smaller but re-adds priming padding at every edge, so keep wav for anything that will be joined again or drives lip-sync.

M4A is a compressed container, so it is a delivery format. If your pipeline later cuts or joins the file, start from wav and encode last.

The Sume detach request

Import the video with POST /v1/media-imports, then send the media.sume.com URL. For a transcript use channels: "mono" and sample_rate: 16000, which the docs call the STT shape. Source video can be up to 1800 s, and output is capped at 900 s, so a longer track needs a range.

A source without an audio track fails with detach_source_has_no_audio, so check probe.has_audio with video inspect before you pay.

• format: wav or mp3.
• range: { start, end? } in seconds.
• channels: source or mono.
• sample_rate: 16000, 44100 or 48000.

Choosing between them

Use Mux's M4A when the goal is playback or a download for a listener. Use Sume's wav when the goal is processing: transcription, splitting into ranges with timeline audio, or driving a render. If you need a smaller delivery file afterwards, mp3 from detach is available but should be the last step in the chain.

What this means for a pipeline

If your video is hosted on Mux and you only need a listening copy, its audio-only rendition is the direct route. If you need a transcript from a video that is already on Sume, detach it at 16000 Hz mono and send that file. The two services do not share files, so pick the one that already holds your video.

Mux says it does not upscale, and skips renditions that would need it. Sume's detach has no such concept for audio, since it extracts rather than resizes.