Audio & display settings

Choose the input and output independently, set stereo pairs and calibrate the listening route. The Configurator prepares boot hardware and the sample-rate ceiling.

Live Pi · October 2026

Open Settings on the Live dashboard at http://PI-ADDRESS:8001/. These settings control the listening session. The Configurator on port 8003 prepares the appliance, preferred sample rate and Audio HAT driver.

Jump to devices and stereo pairs, example connections, buffering, silence measurement, click families or display.

Independent input and output devices

Current Live Pi Settings with independent input and output device selectors and stereo pairs
Actual b1-28 Settings. Both saved selectors happen to use USB Audio CODEC here; they are independent choices. Values are examples, not universal defaults.
Input device
The capture source feeding restoration, grading and calibration. A capture-only USB phono preamplifier or audio interface can be selected without requiring it to have an output.
Input pair
The stereo connectors receiving your phono-stage signal. On a multichannel interface, select the intended pair; processing remains stereo.
Output device
The playback destination, selected independently. It can be the input interface's output, a second USB interface/DAC, a detected output HAT or the available Pi 4 analogue jack.
Output pair
The stereo connectors feeding the amplifier or listening system. Input and output pairs need not share the same device or channel numbering.

The menus use stable device identities instead of Linux card numbers that can change after boot. Only devices capable of the relevant direction appear. Pair choices reflect the device's topology at the negotiated rate, not a promise that every pair exists at every rate.

Automatic selection

With both directions set to Automatic, Live prefers a single duplex interface where possible. With one device chosen explicitly, its Automatic companion prefers the same card when that direction is available; otherwise it resolves independently. An explicitly selected but absent device is reported unavailable rather than silently replaced.

  1. Connect the intended devices and configure any required Audio HAT driver first.
  2. Open Settings, choose Input device and Input pair, then choose Output device and Output pair.
  3. Allow any running audio-session restart to finish, then check the dashboard's actual route and format. A changed device/pair can briefly interrupt audio; it does not reboot the Pi or take down the WebApp.
  4. Recheck hardware input gain and calibration. A route change invalidates route-specific measurements.

Changes are saved on the appliance. Stop before unplugging a device. Repeating the same selection does not intentionally restart audio. Device selection does not replace your restoration profile, Auto, NR or grading policy.

Example audio connections

InputOutputSetup
Duplex USB interfaceThe same interfaceChoose its input/output pairs; one shared hardware clock.
USB capture interface or phono preampSeparate USB DAC/interfaceSelect each direction explicitly; both must support a common native rate.
USB capture interfacePi 4 analogue jackWhen enabled and detected, select bcm2835 Headphones under Output device and connect the jack to a suitable amplifier line input.
USB capture interfaceOutput-only Audio HATPrepare its exact driver in the Configurator, reboot if required, then select the detected output in Live.

The current Pi 4 output menu includes bcm2835 Headphones. Without an active HAT, its jack is prepared at +4 dB ALSA. A configured HAT normally disables that jack. Pi 5 has no analogue jack and the standard image keeps HDMI audio disabled. A DAC-only HAT does not create an input.

Output type matters: amplified HAT speaker terminals must never feed a line input; digital outputs need a compatible receiver or DAC. Follow the board's wiring and power instructions.

Common rate, independent depth and clocks

Both devices must support the same nominal sample rate under the Pi's platform ceiling. Start at 48 kHz; separate devices do not imply arbitrary conversion from a 44.1 kHz input to a 48 kHz output. The Configurator selects the rate ceiling; Live shows the negotiated result.

PCM depths are chosen independently. A 24-bit input can feed a 16-bit-only output without reducing the processing graph's capture precision; a 16-bit-only input remains supported. The footer displays input/processing depth, not both endpoint depths. Consult diagnostics and native audio-format logs for playback details. Distinct device clocks are compensated automatically in final playback. A same-card route bypasses that drift correction.

Clock correction applies after processing and gain to the selected listening signal, not to the raw input, grading or NR learning timeline. The output meters describe software PCM before final clock correction and hardware gain, not the signal measured at the speaker or jack. Check headroom and listen for distortion; software meters do not certify analogue clipping behavior.

Playback buffering

Low latency / Balanced / Safe target approximately 500 / 550 / 600 ms of total software latency. Driver constraints, independent-device queues and downstream hardware can add delay. Safe is a sensible starting point. A buffering change applies at the next Start.

Disable parallel direct monitoring if it mixes undelayed input with the processed output: this can produce an echo or hollow sound. Clock correction keeps independent devices paced together; it is not zero-latency playback.

Noise capture and side-ending silence

NRNG maximum capture level
The loudest eligible noise for local profile capture; −55 dBFS is the usual starting point. Raising it can admit quiet music. This is not the silence threshold.
Silence duration / Silence level
Define sustained quiet input that ends a side. Calibrate for the powered input chain. Use raw input RMS, not output peaks or lowest NR capture RMS.
Quiet passages
Start below and Maximum increase govern extra automatic click/crackle correction in quiet passages. They do not change the displayed media grade or NR limits.

Measure silence (10 s)

  1. Keep the selected audio session running. Lift the stylus or stop the source while leaving the input chain powered; do not use quiet music as the reference.
  2. Choose Measure silence (10 s) in the Audio card and wait for the final result.
  3. Review raw Input RMS, Noise P95 and the suggested threshold. Guidance uses the 95th-percentile raw stereo RMS plus 6 dB, bounded to the allowed threshold range.
  4. Choose Apply silence suggestion only when the measurement is suitable. Nothing is applied automatically; digital-zero or invalid/insufficient input can suppress a suggestion.

This passive ten-second measurement does not suspend restoration or change NR capture. Its result becomes invalid after a route/format change, restart or new measurement. The separate balance/system/surface-noise workflow is explained in Input calibration.

Click Families and processing limits

Current Click Families and Interface settings on Pi 4, showing the ordinary-click restriction
Greyed remembered checkboxes do not mean every family is active. This Pi 4 runs the ordinary-click policy.

The five families are Ordinary clicks, Quiet stereo tails, Quiet channel tails, Short quiet impulses and Compound impulses. Automatic by sample rate uses the platform policy; the checkboxes retain the separate Custom selection.

Pi 4 always uses ordinary clicks only, at a maximum of 48 kHz. On Pi 5, Automatic enables all five families at 44.1/48 kHz and ordinary clicks only at 88.2/96 kHz. Custom selection can override the Pi 5 Automatic policy, but additional high-rate families may exceed real-time processing headroom.

Fully Stop audio before editing families; input-silence pause is not Stop. Apply the remembered choices on the next Start. Selecting no families bypasses DeClick only: DeCrackle, NR and grading remain separate.

Interface and listening-room display

Meter colour and Meter display choose the bars' appearance. Full-screen meter alignment aligns meters to the artwork's lower edge in landscape. Clear Shazam at 100% and Clear after control retention of a completed recognised track. Clock screen saver and Clock delay control the idle display.

Correction profiles

Current Mid AutoNG profile editor with separate DeClick, DeCrackle and NR columns
Current profile editor; it is not the former increasing maximum-repair table.

Select a preset or edit the independent DeClick/DeCrackle AutoNG sensitivity curves and choose Apply as Custom. Built-in sensitivities decrease from Mint to Poor; the two stage scales are independent. Custom values accept 0–100%, with 0 bypassing the corresponding stage. NR limits are independently non-decreasing and can be entered from 1–20 dB; effective S150/S200 operation remains bounded by the engine's 12 dB maximum.

Ordinary settings save automatically, while the profile table has its own Apply action. Restore defaults returns to the default Mid table. See Restoration and grading for listening guidance.