Internal + Public

The AI-Engineer

The AI-Engineer is a chat assistant built into the DAW that knows your project and can do real work - answer questions, generate music and SFX, mix hands-free (mute, volume, pan, effects, automation), replace a section of a track with new AI audio, and batch-export sample packs. Everything it changes is undoable.

Updated 2026-07-26 Music DAW dawaiassistantai-engineermixinggenerationexportsample-packautomationvoice

What it is

The AI-Engineer is a conversational assistant that lives in the DAW's right panel, on its own tab next to the Effects Rack. It understands the DAW and it is aware of your current project (tempo, tracks, levels), and it can take real actions on your mix. You talk to it in plain language.

It is free to chat with (generating music or SFX still costs credits, the same as doing it by hand), and every change it makes to your project is undoable with Ctrl+Z.

Opening it

In the right panel, click the AI-Engineer tab (next to Effects Rack). The effects rack stays put behind the other tab. A quick-start greeting appears the first time, along with a few suggestion chips.

What it can do

1. Answer questions and read your project

2. Generate music and sound effects

A live progress card shows when a real generation is running. Generation needs a signed-in account and credits. See The Library.

3. Mix and edit, hands-free (all undoable)

Refer to tracks by their name and just ask:

If a track name does not match, it tells you instead of guessing. It cannot move or trim individual clips yet; for that, use the timeline controls (see Tracks and clips).

You can also give it a goal and let it work out the details across every track. For example, after loading a stem-separated song whose tracks are named like "Leverage wav Guitar mp3", "Leverage wav Bass mp3", and so on, just say:

"Rename the tracks with only the instrument name."

It renames them all in one go - Guitar, Bass, Vocals, Drums, Piano, Other - each with its own confirmation, and the change shows in the track headers and the mixer. This is handy for tidying up imported stems before you build a sample pack.

A starting mix in one step. Ask for a genre or style and the AI-Engineer sets sensible starting levels across every track at once:

"Please set the track levels for a Pop style mix."

It reads each track's role, makes the lead vocal the reference so it stays present, and sets every other track relative to and below it. Crucially, it leaves headroom - keeping the loudest element off the ceiling so the combined mix peaks below 0 dB rather than clipping. It applies as a single undoable step (one Ctrl+Z reverts the whole balance). This is a starting point to build from, not a finished master, so tweak to taste from there. It works for other styles too ("a lo-fi mix", "a rock mix").

Save your own mix presets. The style mix above comes from the AI-Engineer's general knowledge. Once you have tuned the balance the way you like it, you can save it as your own named preset and recall it exactly on any project:

"Save this as my Pop mix."

That snapshots every track's current volume and pan under the name you give it. Later, on the same or a different song, just say:

"Set my Pop mix."

and it applies your saved levels to the matching tracks in one undoable step - your exact numbers, not a fresh guess. This is deterministic: what you saved is what you get. The AI-Engineer knows which presets you have saved and will use them when you refer to "my" mix by name; ask for a plain style with no saved match and it falls back to building fresh starting levels. You can also say "delete my Pop preset" to remove one. Presets are stored in your browser, so they stay with the machine you save them on.

Save and recall whole mix versions. A preset is a reusable level template; a version is a complete snapshot of this song's mix - every track's volume, pan, mute, and solo and all the effect chains, on the tracks and the master. Use versions to try alternate mixes and jump between them:

"Save this as Mix A." ... (you try a brighter, wetter take) ... "Save that as Mix B." ... "Go back to Mix A."

Saving captures the current mix exactly; recalling restores the whole thing - levels, mute/solo states, and every effect with its settings - in one undoable step. Leave the name off ("save a version") and it auto-numbers Version 1, 2, 3. Versions are tied to the project you are in, so one song's mixes never bleed into another, and they are stored in your browser. Say "delete Mix B" to remove one. This is the safe way to experiment: snapshot where you are, push the mix somewhere new, and step back any time.

Effect recommendations. Beyond levels, the AI-Engineer can suggest a full effects chain tailored to your tracks and your style. Ask:

"What effects should I add for a Pop mix?"

and it walks each track with specific, reasoned moves - a low-cut and presence boost plus compression and a short plate on the lead vocal, glue compression and EQ on the drums, tightening on the bass, a light room on the guitar, gentle glue and tape saturation on the master bus, and so on. It then offers to apply them, and it does not just insert the effect: it dials in the settings it recommended - the EQ presence lift, the compressor ratio and threshold, the reverb size, the delay time - so the chain lands as a real starting patch rather than defaults. Say "yes, add them all" and it inserts each effect with its settings into the right rack, or pick and choose ("just the vocal compressor and the master glue").

You can also fine-tune any effect afterward by asking, and it adjusts the existing effect in place:

"Shorten the vocal reverb to about 1.2 seconds." / "Back the delay feedback off to 15." / "A touch more compression on the drums."

It keeps every value in the effect's valid range, refreshes the knobs in the rack, and, as always, it is undoable with Ctrl+Z. The dialed-in values are a musical starting point - open the rack and nudge them to taste.

Automation, drawn for you. Anything that needs to change over time is automation, and the AI-Engineer can draw the curve rather than leaving you to click breakpoints into a lane. Ask for the move:

"Fade the intro in." / "Automate a filter sweep into the drop." / "Duck the pad under the vocal." / "Pan the guitar left to right." / "Ride the vocal level up in the chorus."

It automates a track's volume, its pan, or any parameter of an effect already on that track (add the effect first, then ask for the sweep). It writes the breakpoints into the track's automation lane, so you can open the lane and drag them afterward, and both playback and your export follow the curve. Values are clamped to the parameter's real range. Ask it to clear a curve and it deletes that parameter's automation. See Mixing for the lane itself.

4. Replace part of a track with new AI audio

The Replace Audio chip re-renders a section of a track instead of the whole thing. Use it to redo a chorus, rebuild a weak bridge, or drop a brand new section into an arrangement.

Before you click it, mark what you want changed: drag a loop range on the ruler, or click the clip you want to work on. The card reads that selection and the track it belongs to, and shows both back to you so you can check the target before spending anything.

Click Replace Audio in the AI-Engineer panel and pick one of three operations:

Operation What it does
Replace this part Regenerates the selection, using the audio on either side as context so key and tempo stay continuous. This is the default when something is selected, and the strongest of the three.
Restyle it The same part reimagined. Keeps the underlying structure and timing, changes the style, energy, or instrumentation.
New section Brand new material with nothing to preserve. This is the only one that does not need a selection.

Then describe what you want in the text box ("harder-hitting drop with a heavier snare") and click Go.

It costs 60 credits per render on paid plans, stated on the card before you commit, and a render can take a couple of minutes.

How the result lands. Nothing is overwritten. The DAW duplicates the source track, carves the selected region out of the duplicate, drops the new audio into that hole, and places the duplicate directly below the source. That means you can mute one against the other and A/B the new section in full context, with the rest of the arrangement still playing. Keep the take you prefer and delete the other. Like everything else here, it is undoable with Ctrl+Z.

If you pick Replace or Restyle with nothing selected, it tells you to select a region first rather than guessing.

5. Batch-export mixes and sample packs

This is a highlight for sample-pack creators. Ask for several full-song WAV mixes at once and each downloads as its own file:

"Export full WAV mixes of: drums muted, bass and drums only, and vocals only."

Each mix mutes or solos the right tracks, renders through the real export engine (your faders and effects included), and saves a WAV named project_mix.wav. Your live mix is restored afterward.

It can also render with quality options - format, bit depth, sample rate, MP3 bitrate, loudness, and stems - the same choices as the Render dialog. Just say what you want:

"Render a 24-bit 48k WAV." / "Bounce a 320 MP3." / "Export 24-bit FLAC stems." / "Render and get me to -14 LUFS."

Defaults, if you do not specify, are a 24-bit / 48 kHz WAV with no loudness change. See Export and render for what each option means.

Setting an export folder

By default, exports go to your browser's Downloads folder. To send them to a specific folder instead (great for keeping a sample pack together):

  1. Click the Export folder button in the AI-Engineer panel.
  2. In the folder picker, navigate to and select (or create) your folder, for example a "Leverage-Sample-Pack-144bpm" folder, and allow write access.
  3. The button lights up with the folder name. For the rest of the session, all exports - the AI batch export and the main Export button - save straight into that folder.

Browsers do not let a web page write to a typed absolute path for security, so you choose the folder through the picker rather than typing it. This uses the File System Access API and is available in Chrome and Edge; on other browsers exports fall back to Downloads.

Foundational Samples

Foundational Samples is a one-click sample-pack preset. Click the gold Foundational Samples chip (or type "export Foundational Samples"), and the AI-Engineer proposes this set, mapped to your actual track names:

It confirms first: it lists exactly what it will export and waits. Reply "yes, all of them" to export the batch, or exclude items, for example "exclude the piano stem", and it drops those and exports the rest. Set an export folder first so the whole pack lands in one place.

Name your engineer

Click the Name chip and give the assistant whatever name you want, up to 24 characters. The name replaces "AI-Engineer" on the panel tab and in the chat input placeholder, so the DAW reads as your studio.

This is a subscriber feature. On the free tier the chip shows a Subscribe link instead of the rename card. Save an empty name to go back to the default. The name is stored in your browser, so it stays with the machine you set it on.

Talk to it, and let it talk back

Two buttons sit next to the chat input:

Both use the browser's built-in speech engine, so there is nothing to install and nothing extra to pay for.

Browser support. Reading replies aloud works in every current browser. Voice input needs Chrome, Edge or Safari; it is not available in Firefox. Where it is missing the buttons are simply not shown, so nothing breaks.

About the voices. The list comes from your browser and operating system, not from us, so it differs per machine. Voices marked with an asterisk are network voices and are usually the more natural ones. The DAW picks the best available English voice for you rather than the system default, which on Windows is a noticeably flat robotic voice. Pick a different one any time.

Tips

Where to go next