add soundtrack

[jeremymanning]

include an animation flag that will (optionally) add a dramatic soundtrack (or a user-selected audio file) to animations.

eventually we could automatically generate an audio soundtrack using the data. 🎶

[jeremymanning]

Status in the 1.0 update (PR #272 #272): unchanged — no audio/soundtrack support anywhere in hypertools/plot/animate.py or hyp.plot's signature; no audio-related dependency in pyproject.toml.

How this could be built on the 1.0 architecture: hypertools/plot/animate.py's _save_animation already requires ffmpeg for any non-gif/svg/png animation export (animation.writers["ffmpeg"]), and ffmpeg is confirmed present on this system (/opt/homebrew/bin/ffmpeg, v8.1.2). That means user-supplied-audio muxing needs no new dependency: after _save_animation writes the silent .mp4, shell out to the already-required ffmpeg binary to mux in an audio track (ffmpeg -i video.mp4 -i track.mp3 -c:v copy -c:a aac -map 0:v:0 -map 1:a:0 -shortest out.mp4). Concretely: add a soundtrack=<path> kwarg threaded from hyp.plot() down to _save_animation, restrict it to container formats that support audio (mp4/mov/avi — not gif/svg/apng, which have no audio channel), and raise/warn if soundtrack is set for those.

Alternatives / prior art:

• moviepy/pydub would simplify the muxing call into a couple of Python lines, but neither is installed today and both would add a new dependency purely to wrap the same underlying ffmpeg call hypertools already relies on — a direct subprocess call to the existing ffmpeg dependency is more consistent with the current architecture and avoids the extra install.
• For the "auto-generate a soundtrack from the data" stretch goal (sonification): domain-specific prior art exists (e.g. NASA's astronify for data sonification, or simple pitch-mapping via numpy+scipy.io.wavfile with no new heavy dependency), but this is materially harder — it needs a design decision about what data feature maps to what audio parameter, which is subjective and dataset-dependent.

Note: this is a nice-to-have relative to hypertools' core visualization/analysis work; no evidence of renewed demand since filing.

Tagging as medium effort (user-supplied audio file muxing via the already-required ffmpeg binary); the auto-sonification stretch goal is separately high effort and would warrant its own issue. Leaving open for consideration in a future release.

[jeremymanning]

Leave this open; punt for a future release