AutoEQ CLI

The autoeq binary optimizes parametric EQ filters to flatten a speaker or headphone’s frequency response. It can fetch measurements directly from spinorama.org or read local CSV files.

Quick Examples

Optimize a speaker from spinorama.org (7 filters, COBYLA):

autoeq --speaker "KEF R3" --version asr --measurement CEA2034

Optimize headphone from a local CSV with 5 filters and a custom target curve:

autoeq --curve headphone.csv --target harman-over-ear-2018.csv --loss headphone-score -n 5

Use Differential Evolution for a global search:

autoeq --speaker "Genelec 8030C" --version asr --measurement CEA2034 --algo autoeq:de -n 7

Input Sources

From spinorama.org (API)

Use --speaker, --version, and --measurement together:

autoeq --speaker "Yamaha NS-10M" --version asr --measurement CEA2034

For CEA2034 measurements, optionally choose which curve to optimize against (default: Listening Window):

autoeq --speaker "JBL M2" --version asr --measurement CEA2034 --curve-name "On Axis"

From a CSV File

Provide a frequency-response file with --curve. Columns: frequency,spl:

autoeq --curve my-speaker.csv

Add a custom target curve with --target (same format):

autoeq --curve my-speaker.csv --target target-curve.csv

Core Parameters

Flag	Default	Description
-n / --num-filters	7	Number of IIR biquad filters
--min-db	1.0	Minimum filter gain (dB)
--max-db	3.0	Maximum filter gain (dB)
--min-q	1.0	Minimum Q factor
--max-q	3.0	Maximum Q factor
--min-freq	60	Minimum filter frequency (Hz)
--max-freq	16000	Maximum filter frequency (Hz)
--sample-rate	48000	Output filter sample rate (Hz)

Loss Functions

Select with --loss:

Value	Use case
speaker-flat (default)	Flatten speaker response to 0 dB
speaker-score	Optimize Harman/Olive preference score (adds bass boost)
headphone-flat	Flatten headphone response toward a target
headphone-score	Target Harman headphone curve
drivers-flat	Multi-driver loudspeaker crossover optimization

Optimization Algorithms

Select with --algo. Use --algo-list to see all options.

Value	Type	Notes
nlopt:cobyla (default)	Local	Fast, reliable for ≤10 filters
autoeq:de	Global (DE)	More thorough; use with --strategy
nlopt:isres	Global	Slower but explores wide parameter space
nlopt:nelder-mead	Local	Good for fine-tuning

For Differential Evolution, configure the search strategy with --strategy (default: currenttobest1bin). List available strategies with --strategy-list.

PEQ Filter Models

Control filter structure with --peq-model. Default: pk (all peak filters).

Model	Structure
pk	All peak/bell filters
ls-pk-hs	Lowshelf + peaks + highshelf
hp-pk-lp	Highpass + peaks + lowpass
free	Any filter type per band

List all models with --peq-model-list.

Presets

Named presets set all optimizer parameters at once. Individual flags override:

Preset	Description
quick	Fast result, fewer evaluations
balanced	Good quality/speed tradeoff
max-quality	Maximum optimization effort
score	Optimized for speaker preference score

autoeq --speaker "KEF R3" --version asr --measurement CEA2034 --preset balanced

Output

By default, autoeq prints optimized PEQ filters to stdout in a human-readable format.

Save a plot of the result (requires plotly_static feature):

autoeq --speaker "KEF R3" --version asr --measurement CEA2034 --output result.png

Reproducible Results

Set --seed for deterministic optimization:

autoeq --curve speaker.csv --algo autoeq:de --seed 42