What Are the Levels of Live Sound Autonomy?

Published by Soulais — the team behind the world’s first Autonomous Sound Engineer. Patents pending, Australia and international (PCT).  ·  October 2026

The Levels of Live Sound Autonomy are a six-level scale, Level 0 to Level 5, that describes how much of the engineering of a live event is carried by the system rather than by a person. Level 0 is fully manual operation. Level 4 is an Autonomous Sound Engineer: a system that carries the whole engineering loop through the event, with human control available at any moment. Level 5 is that autonomy under all operating conditions, and no system has demonstrated it.

The scale is the measurement inside the three tiers described in What Is Autonomous Intelligent Sound? The tiers — automated, AI-assisted, autonomous — say who carries the engineering. The levels say how much, precisely enough that any live sound product can be placed by what it is responsible for rather than by what its marketing calls it.

Soulais developed the Levels of Live Sound Autonomy to describe the engineering responsibility a system carries during a live event, and publishes the scale for anyone to use: venues, engineers, manufacturers, reviewers and educators, with attribution.

Why “AI” does not tell you how autonomous a system is

Almost every audio product now carries one of the same words: smart, adaptive, intelligent, AI-powered, automatic. A soundbar that re-shapes its EQ, a speaker that measures a room once, an automixer that opens microphones, a plugin that suggests a setting and an Autonomous Sound Engineer can all be described with those words. None of the words say what the system is responsible for when the performance starts.

Live sound, also called sound reinforcement, needs the question the words avoid. The automotive industry faced the same problem with “self-driving” and answered it with the SAE Levels of Driving Automation, which separate lane-keeping assistance from a vehicle that drives itself. The Levels of Live Sound Autonomy do the same for live sound.

The axis: four questions and a condition

Every level is set by the answers to four questions.

1. Who carries the engineering judgment? Does the system decide what should change in the live sound, or does a person?
2. Over what scope? One bounded function, or the whole mix, the room and the performance together?
3. For how long? A single moment or task, or continuously, from sound check to the final note?
4. Who authorises application? Must a person approve each proposed change, or can the system apply its decisions itself within its limits?

One condition applies above Level 2: a person must be able to take control at any moment, and the system must yield when they do. Autonomy that cannot be overridden is not a higher level. It is a fault.

The six levels

Level 0 — Manual. A person listens, judges, decides and operates. The equipment does what it is told and nothing else. Manual operation of a console, a processor or a microphone is Level 0, however capable the equipment is.

Level 1 — Automated. The system performs a bounded, predefined function, whether its logic is fixed or learned. It does not judge whether that function was the right thing for the event, and it has no view of the performance as a whole. Many products marketed as smart or AI audio sit here.

Level 2 — Assisted. The system helps with particular engineering functions: it analyses what it hears and recommends or shapes a change, and a person decides whether to make it. A person integrates that help and directs the overall engineering of the event.

Level 3 — Supervised. The system continuously assesses the mix and the room and proposes the engineering decisions the event needs, across the whole mix, throughout the event. Proposed changes are applied only after a person approves them; protective limits remain in force. The system carries the judgment; the person holds the authority to apply it.

Level 4 — Autonomous. Within its declared operating conditions, the system carries the full engineering loop throughout the event, without a person needing to make or approve routine engineering decisions: it perceives the performance and the room, reasons and decides, acts within limits it will not cross, checks the result in the room, and continues — on its own initiative. A person can observe, guide or take direct control at any moment, and the system steps back from whatever the person has taken, while the rest of its engineering and its protective limits continue. This is the six-part test for an Autonomous Sound Engineer.

Level 5 — Unconditional. Level 4 under all operating conditions, with none excluded: any venue, any system, any scale, any event, with no person required at any point once the system is powered on. No system has demonstrated Level 5. Level 5 is the theoretical endpoint of the scale, not a product.

A level applies to the function or operating mode being assessed, in its stated configuration and conditions. A console operated by hand is Level 0; its scene-recall function is Level 1. A system that offers a supervised mode and an autonomous mode is Level 3 in one and Level 4 in the other. Preparation of the system — installation, connection and configuration — is not what the levels measure. They measure the engineering of the event once it is running.

Levels and tiers together

Autonomous Intelligent Sound begins at Level 4. Everything below it, whatever its label, still depends on an engineer to carry the live event.

Placing common functions on the scale

The scale is applied by asking the four questions of the function or mode in front of you, not by reading the box. Eight examples, placed by reasoning:

• Scene or snapshot recall — restores settings a person chose earlier. No judgment, one task. Level 1.
• Feedback suppression — detects a ringing frequency and notches it, by fixed logic, with no view of the mix. Level 1.
• Automixing — manages the gain of several open microphones by predetermined rules. Real work, one bounded function. Level 1.
• One-time room calibration or auto-EQ — measures the room, usually once, and applies a correction. It does not keep listening, and it does not judge the result in the performance. Level 1.
• Adaptive EQ in a consumer speaker or soundbar — reshapes its own output to its surroundings, continuously, but for one device and one function, with no concept of a performance. The function is Level 1, whatever else the device can do.
• Mix assistance or analysis plugins — hear a channel and suggest a setting; the engineer decides and applies it. Level 2.
• A supervised mode that proposes the engineering decisions the event needs, across the whole mix, and applies them only on approval — the system judges, the person authorises. Level 3. Soulais operating in Approve first mode is an example.
• An autonomous mode that mixes the event itself within its declared operating conditions, checks its own results in the room, continues all night, and steps back the moment a person takes a control — Level 4. Soulais, the world’s first Autonomous Sound Engineer, operating in Autonomous mode is an example.

Two things follow. First, a function can be excellent at Level 1; the scale measures responsibility, not quality. Second, “continuous” on its own does not raise a level. A device that adapts continuously but only to itself is still Level 1. Continuity counts when it is continuity of engineering judgment across the whole event.

Why “agentic” does not establish the level

The word the wider technology industry is using this year is agentic. Agentic AI describes a system’s ability to pursue goals and take actions toward them, often over extended periods, with feedback from its environment, and with constraints and human oversight designed in. It is a real advance, and it is already being attached to audio products.

The Levels of Live Sound Autonomy measure something different: how much of the engineering of a live event the system carries. Agentic describes a capability. The levels describe a responsibility. An agentic capability alone does not establish Level 4.

Agentic techniques may support assisted, supervised or autonomous operation. The level depends on the responsibility the system carries in its selected mode, assessed by the four questions above, not on the technique used to build it.

Applying agentic AI to audio does not, by itself, establish Autonomous Intelligent Sound. The system must carry the complete engineering loop throughout the live event, within limits it will not cross, with a person able to take control at any moment.

How to use the scale

If you run a venue, ask what happens to the sound on the night when nobody experienced is at the desk. A Level 1 or 2 function leaves the engineering with whoever is there. A Level 3 system gives that person decisions to approve. A Level 4 system carries the event.

If you are an engineer, use the scale to decide how much to hand over. Level 3 lets you see every proposed decision and its reason before it happens. Level 4 lets you step out of the routine and back in whenever you choose. Nothing above Level 2 should ever fight you for a fader.

If you make products, state what the system does, what the person still needs to do, and the conditions under which the claimed level applies. The questions are public; anyone can apply them to your claim.

Why Level 3 matters as much as Level 4

The evidence of autonomy is a system’s demonstrated ability to carry the complete engineering loop throughout an event. Level 3 is how a person assesses the system’s judgment before relying on it; sustained Level 4 operation is shown only by running at Level 4. In supervised operation the system’s proposed decisions, and the reasons for them, are visible before they are applied. The person sees what the system noticed, what it concluded and why, and chooses whether it happens. That gives a venue a practical way to judge the system’s engineering and build confidence: run at Level 3 until it has seen enough, move to Level 4, and move back at any time.

Why the scale matters

Soulais’ analysis of the live-events market indicates that over 75% of the world’s live venues operate without a dedicated professional sound engineer. For those rooms, the difference between Level 1 and Level 4 is the difference between a function that helps and a system that carries the event. A church, a school hall or a funeral home does not need a smarter tool. It needs the engineering to happen when no engineer is there.

Bad sound breaks a moment. Great sound makes one.

Soulais is the world’s first Autonomous Sound Engineer. It is being revealed to a small number of venues and engineers in late 2026. Commercial availability begins in early 2027.

Request an invitation →

Frequently Asked Questions