Understand Every Echo Live Setting

Rescans microphones and outputs connected after Echo opened.

Recommended: Use this before restarting Echo when a new headset or microphone is missing.

The physical microphone Echo processes.

Recommended: Choose your real microphone, not the virtual cable.

The virtual route Echo sends to Discord, games, OBS, and other apps.

Recommended: Use the automatically detected cable unless you have an advanced routing setup.

Turns local headphone monitoring on or off.

Chooses where your local monitor is played.

Recommended: Use headphones or speakers you actually listen to, not the virtual cable.

Changes only the local monitor, from 0% to 200%.

Turns microphone cleanup on or off.

Lets Echo choose an available cleanup backend.

The normal general-purpose cleanup default.

Recommended: Start here unless your microphone has a repeatable problem with it.

An alternative cleanup path that can behave differently with room noise and quiet speech.

A hardware-dependent option shown only when the required support is available.

Changes how quiet input must be before Echo treats it as silence.

Reduces low-frequency rumble such as desk vibration and handling noise.

Helps when speaker output can re-enter the microphone.

Recommended: Headphones usually give cleaner and more predictable results.

Shows the backend used by the currently loaded model, such as CPU, DirectML, CUDA, or CoreML.

Checks whether Echo can start GPU voice processing on the selected adapter.

Chooses GPU or CPU for the next model load.

Recommended: Reload the voice after changing this setting.

Chooses Auto or a specific detected GPU.

Recommended: Use Auto unless it selects the wrong device or support asks you to test one.

Shows inference features compiled into the current app build.

The large value is the smoothed time Echo spends processing recent AI blocks.

The processing time for the latest completed block.

The time budget available for each converted block.

Chunk window minus processing time. Small or negative headroom means the current mode may be too aggressive.

Blocks that did not finish inside their processing window.

Separate local estimates for Virtual output and Hear yourself when enough timing data exists.

Follows input level through speech, quiet audio, and silence.

Recommended: This is the normal default.

Makes an explicit decision about whether the current input is speech or silence.

Uses the microphone preprocessor's speech decision. Available only with Gate-based processing.

Uses signal level for the gate decision. Available only with Gate-based processing.

Offers Legacy / Off, Smooth fade, Smooth fade + reset, and Aggressive cleanup for Gate-based processing.

Controls the primary conversion window. Smaller values can reduce delay but leave less time for inference.

Provides surrounding audio history to help continuity.

Controls overlap between converted blocks. More overlap can smooth boundaries but soften transients.

Selects bundled RMVPE or CREPE pitch tracking.

Recommended: Keep RMVPE unless you are comparing a repeatable pitch-tracking problem.

Expose extreme minimum-delay, CPU-safe, and maximum-quality combinations.

Changes the app interface language. Some newer specialized areas may still contain English while translations catch up.

Changes the app look without affecting audio or performance.

Shows CPU, RAM, GPU, driver, operating system, disk, free space, and compatibility warnings.

Copies the hardware and selected-backend summary for a technical report.

Starts Echo's browser authentication flow.

Refreshes account entitlement after a purchase or account change.

Replays first-run setup without reinstalling Echo.

Opts out of optional product analytics.