# shine — fixed-point MP3 encoder

Vendored verbatim from [toots/shine](https://github.com/toots/shine) at commit
`ab5e3526` (2023-01-01). See `UPSTREAM_VERSION.txt`.

Treat this as third party, the same way `lib/ST7305_U8g2/` is treated: fix bugs
upstream-style and keep local changes minimal and commented. **There are
currently no local changes** — `src/` is a straight copy of upstream's
`src/lib/`.

## Why it is here

The ESP32-S3 has no hardware audio encoder, and raw 16-bit PCM from the ES7210
is 32 kB/s — about four minutes into the whole PSRAM heap. Shine is fixed-point
and skips psychoacoustic analysis entirely, which is what makes real-time
encoding possible on this part at all. It is the encoder the ESP32 audio
community has settled on for exactly this reason.

Only the parts of upstream needed to build the library are copied: the autotools
scripts, the JS/Android bindings and the command-line front end are all left
behind. Shine needs no generated `config.h`, so it compiles as-is.

## Sample rates

Upstream supports MPEG-2 LSF, so **16 kHz is a valid sample rate** (`tables.c`
lists 16000 under MPEG-II). Some older forks are MPEG-1 only, which is where the
widely repeated "shine cannot go below 32 kHz" claim comes from. It does not
apply to this copy.

## Licence

Shine is **LGPL-2.1**, which is more restrictive than everything else in this
repository. Statically linked into a firmware image, the LGPL's relinking
obligation is only really satisfiable by publishing the object files or the
source. That is not a problem for a personal device whose source is already
public, but it is worth knowing before this firmware is distributed as a binary.
`COPYING` is upstream's licence text, unmodified.