Is this video stereo or mono? Read probe audio_channels first

To find out whether a video carries stereo or mono sound, run Sume video inspect and read probe.audio_channels (an integer, or null), plus audio_sample_rate and audio_codec. These come from the same ffprobe pass as the video fields, and has_audio tells you whether there is a track at all.

Check rather than assume. A mono dialogue clip and a two-channel music-bed clip both come out of one detach call, but they want different settings.

Choose detach settings from the probe

Why the defaults matter

Audio detach returns sample-exact wav (pcm_s16le) by default, and channels defaults to source, so a stereo track stays stereo. Setting channels: "mono" downmixes. sample_rate takes only 16000, 44100 or 48000, and 16000 with mono is the speech-to-text shape.

The output cap is 900 seconds, and the source cap is 1800 seconds. A whole track longer than 900 seconds needs a range. Public rate is $0.01 per job.

Two lines of logic

Read the probe, then decide.

probe = result["probe"]
if not probe["has_audio"]:
    raise SystemExit("no audio track: skip detach")
body = {"video_url": url}
if probe["audio_channels"] and probe["audio_channels"] > 1:
    # keep stereo for a mix; downmix only for speech-to-text
    body["channels"] = "source"
print(body)

Detach once, then split many ranges later with timeline audio, rather than detaching the same video repeatedly.