sweeptune

Minimal macOS CLI that measures your listening room with a log sine sweep and tells you what to enter into the Fosi Audio S3 10-band EQ (31.5 Hz – 16 kHz, ±10 dB, 0.5 dB steps) to reach a neutral-but-pleasing, Harman-like in-room curve. It also reports reverberation (RT60), early reflections and room modes, because EQ cannot fix echo.

Install

brew install uv
uv tool install sweeptune     # from PyPI, puts `sweeptune` on your PATH
sweeptune --help

From a clone instead: uv sync, then prefix the commands below with uv run.

Setup

Select the Fosi S3 as the Mac's sound output (Control Center → Sound → AirPlay) before measuring. Device names can be overridden in an optional sweeptune.toml (see src/sweeptune/config.py for all keys):

output_device = "S3"        # name fragment of the AirPlay output
input_device = ""           # name fragment; empty = mic profile default
max_boost_db = 4.0
shelf_db = 6.0              # bass shelf of the harman-like target

Workflow

uv run sweeptune doctor                         # routing, mic inputs, noise floor
uv run sweeptune balance --live                 # set the amp's balance knob at the seat (L/R level + distance)
uv run sweeptune measure --mic audyssey         # 7 positions, L and R, EQ flat
uv run sweeptune eq suggest                     # table + copy string for the S3
uv run sweeptune plot                           # frequency response vs target + suggested EQ (PNG, opens in Preview)
# enter the values on the S3 (and lower its volume by the headroom shown)
uv run sweeptune verify                         # re-measure, compare to prediction, next correction
uv run sweeptune verify                         # repeat until converged
uv run sweeptune report                         # sessions/<timestamp>/report.md + PNGs

Balance

sweeptune balance --live plays L, R, L sweeps at the listening seat and shows how much louder one side is (500 Hz – 4 kHz, where the stereo image lives) and how much farther one speaker is. Turn the amp's balance knob, press Enter to re-measure, and stop within ±0.5 dB. Fix a distance difference by moving the speaker first; the balance control is for the remaining level difference. Without --live, sweeptune balance reports the same figures from the latest multi-position measurement. The S3 EQ acts on both channels equally, so balance and EQ can be set in either order.

Optional:

• sweeptune eq probe 125 – change one band by −6 dB and learn the S3's real filter Q (saved to streamer.json).
• sweeptune mic relcal iphone17pro --ref audyssey – derive a calibration for another mic at the same spot.
• sweeptune sweep export + sweeptune analyze rec1.wav rec2.wav … – offline path, e.g. recording on the iPhone with an app that offers a measurement (unprocessed) mode.
• sweeptune plot --channels – left and right speaker on a shared level, so a balance offset is visible.
• sweeptune --simulate … – run any command against a synthetic room, no speakers needed.

Microphones

None of these mics has a traceable calibration, so all results are relative. The MacBook jack is CTIA TRRS; if doctor finds no external mic, the Audyssey plug needs a TRS→TRRS mic adapter.

Tests

uv run pytest

Releasing

1. Bump version in pyproject.toml (PyPI never accepts the same version twice), commit and push.
2. Publish a GitHub release whose tag is that version with a v prefix, e.g. v0.2.0. The publish workflow tests, builds and uploads to PyPI via trusted publishing (no token needed).

License

MIT © 2026 Maximilian Schuh. Free to use, modify and redistribute, including commercially, as long as the copyright notice and license text are kept — i.e. with acknowledgement.