Turbo mode in Maestra: direct commands, what they cost and how to switch it off

The short answer: Turbo is a quick decision step that runs before the full model. When your message is a direct command, such as "mute the bass" or "fire scene 2", Turbo picks the action, the target and the value, and Maestra carries it out and reads it back, without a full model turn. Turns that go through the decision step use 1.5× credits. Everything else runs exactly as it does without Turbo.

From Maestra 1.0.5, Turbo is on for everyone, on every plan. You can switch it off in one click, next to the chat box.

What a direct command is

A direct command names one bounded action on one target. Mute this, set that, load this, fire that. The main examples:

• Mixer: "mute the bass", "solo the vocal", "set the Drums volume to 70%", "pan the hats left", "arm the Keys track".
• Transport and tempo: "play", "stop", "set the tempo to 124".
• Clips and scenes: "fire scene 2", "launch the clip on Drums in scene 3", "add a scene".
• Devices: "put a Compressor on 3-Audio", "set the Reverb's decay to 3.5 s".
• Walking back: "undo that".

Creative or open-ended requests are not direct commands. "Write a bassline that fits the chords" is recognised as creative work and goes straight to the full model. A broad request such as "set this session up for mixing" reaches the decision step, which hands it to the full model.

How a direct command runs with Turbo

Without Turbo, every message starts a full model turn. The model reads the context, decides what to do and calls a tool. For a one-knob change, that is a lot of work.

With Turbo, a quick decision step looks at the message first. It chooses from the actions Maestra already knows are possible in your session. If the message is a clear direct command, that choice goes straight to the tool, and Maestra verifies the result by reading it back from Live. If the decision step is unsure, Maestra asks one short question. If the message is not a direct command, it hands the turn to the full model.

The read-back is the same with or without Turbo. The reply reports what Live confirmed.

What it costs

A credit is one exchange with the model. A turn that goes through Turbo's decision step uses 1.5× credits, for the decision and for every exchange in that turn. That includes a turn the decision step hands to the full model. Messages that never go to the decision step, such as a clearly creative request or a follow-up in the conversation, are charged as normal turns, and so is a turn where the decision step failed and Maestra fell back to the full model.

Settings shows the price on the Turbo row: "Uses 1.5× credits." The chat box tooltip says the same. Your balance is in Settings › Activation.

Where to switch it off

Turbo is on by default. There are two places to change it, and they are one setting:

• The Turbo switch in the row under the chat box.
• Settings › Agent › Agent controls › Turbo.

If you had switched Turbo off in 1.0.4 or earlier, updating switches it on once and leaves a note in the chat. Turn it off again and it stays off.

Switching Turbo off always works. Switching it on can be greyed out, with the reason shown:

• "Turbo is not enabled for this licence."
• "Turbo isn't live on the server yet."
• "Turbo becomes available once Maestra can reach your account."

If Turbo stops for the session after the server refuses it, toggling the switch in Settings turns it back on.

Autonomy: Careful, Balanced and Yolo

The same row under the chat box sets how far Maestra goes before asking. Turbo decides whether a direct command skips the full model turn. Autonomy decides when Maestra checks with you first.

• Careful: it asks before anything unclear and shows the plan first. Every change waits for your ok.
• Balanced: it acts when the request is clear and tells you what it assumed. This is the default.
• Yolo: it acts on its best reading and reports after.

Deletes and other destructive changes wait for your ok on every setting. Ctrl+Shift+A (Cmd+Shift+A on a Mac) cycles the three. A change applies to the next action. A confirmation already on screen still waits for your answer.

Performance lock

The row also holds a lock. Turn it on and Maestra blocks live changes while you perform. Turn it off when you are back to building.

When to turn Turbo off

Turbo is worth having on for quick mixer and launch work, where a short turn matters most. You might turn it off if you mostly write longer requests about your session that are not clearly creative, since those go through the decision step and are charged at 1.5× even when it hands them to the full model. With Turbo off, every turn is charged at the normal rate and runs through the full model.

If you are unsure, ask Maestra: "What does Turbo cost?" It answers in the chat.

Summary

Turbo is a quick decision step in front of the full model. A direct command such as "mute the bass" goes straight to the tool, and Maestra reads the result back from Live. Turns that go through the decision step use 1.5× credits. It is on by default from 1.0.5 on every plan, and the switch next to the chat box turns it off. To try a few direct commands on your own set, start on the Get started page.