feat(demo): lay the generated music bed under the demo video

[upgradedev]

The demo film had no music because an unlicensed track is a copyright problem we will not hand a judge. Generating it removes the question instead of dodging it: the bed is as clearly ours as the footage.

minimax-music-2.5 through the same Genblaze path every other asset uses, prompted from the occasion the reel in the video actually picks (anniversary: warm romantic strings, 96 bpm), which is exactly what Occasion.music_style has always documented itself as being for.

Two problems worth naming

No duration control. The bed came back 90.26s against a 169.74s film. demo/mix-music.py loops it by crossfading it into a second copy of itself with a constant-power (qsin) seam, then trims to the film and eases in and out. On a sustained strings pad the seam is not something you can point at.

Narration stays dominant. Both tracks are measured with EBU R128 integrated loudness and the bed is attenuated to sit exactly 20.1 LU under the voice, with amix normalize=0 so the mixer cannot quietly halve both inputs and undo the measurement. The script fails hard if the measured separation drifts more than 0.5 LU.

Integrated loudness is -25.9 LUFS before and after, so the voice is not one decibel quieter than the cut that was already approved.

The picture is untouched

The video stream is copied, not re-encoded, so every footage beat, caption window and cut survives byte for byte. 169.74s, under the 180s cap, ten beats and ten cues.

The gate

The mix is one AAC track, so exactly one audio stream still guards against a stray stream but can no longer tell narration-only from narration-plus-music. check_music adds two bounds, two-sided on purpose (a one-sided "not silent" check would pass a mix that drowns the voice):

• Floor. Narration is built by padding each beat with digital silence, so a narration-only film leaves 44 windows of true dead air at or under -50 dBFS. With the bed under it that count is 1. A rebuild that skips the music step fails here.
• Ceiling. The mixed film must still measure within 1.0 LU of the narration it was built from. A bed mixed anywhere near the voice drags that up and fails.

The bed is pinned by content hash and recorded in demo/cinemory-demo.beats.json with its provider, model and prompt like every other generated artifact.

Docs

The "no music, deliberately" claims in demo/SUBMISSION.md and demo/video-script.md become false on merge and are updated. demo/STATE.md gets a new top entry rather than a rewrite, per its own append-log contract. README.md had no false claim; one clause was added to the demo-video bullet. Badge row, YouTube URL and test-count floors untouched.

Owner action after merge: re-upload demo/cinemory-demo.mp4 to YouTube and repoint the Devpost video link.