🪄 AI Generators, ElevenLabs Sound FX & Smart Melody

Omni DAW features an integrated suite of artificial intelligence and generative algorithms, supporting creators in crafting unique Sound FX, TTS voiceovers, chord progressions, and spectral textures.

1. ElevenLabs Integration: Sound FX & Voice Generation

Omni DAW has a direct connection to the ElevenLabs Sound Generation & Text-to-Speech API:

A. Sound FX Generation

Type a text prompt describing the desired effect (e.g., "Laser blast in deep space", "Dark monster roar", "Tapping on glass transition"), and the ElevenLabs engine will generate a clean, high-quality Sound FX audio file directly into your project!

B. Voice and Voiceover Generation (Text-to-Speech TTS)

• Model and Voice Selection: Access to the latest ElevenLabs vocal models.
• Speech Generation: Transforming text into natural-sounding vocal, voiceover, and narrative lines with emotion and stability adjustments.

C. Application in DAW Modules:

• ScratchDeck Integration: A generated Sound FX sample or vocal can be immediately sent to the ScratchDeck module for live vinyl scratching!
• Convolution Reverb IR Generation: Creating synthetic impulse responses (IRs) for the Conv Reverb effect.
• Sample History: Automatic saving of all generated samples to disk in WAV/MP3 format with the ability for quick reuse.

2. Smart Random Melody Generator (Algorithmic Melodies)

The Piano Roll editor features a Smart Random Melody generator utilizing 12 compositional archetypes:

1. Stepwise Flow: Smooth stepwise motion, arched singing phrasing (Pop / Folk / Singer-Songwriter).
2. Minimal Cell: Short 2–4 note motifs repeated with transposition and inversion (Minimalism / Techno).
3. Classical Arch: Classic phrase architecture with a clear build-up to a climax and decay.
4. Funk Groove: Syncopated rhythm, chromatic neighbor notes, and emphasis on the accented beat (Funk / R&B).
5. Celtic Reel: Very fast stepwise runs and ornaments (Jig / Celtic Reel).
6. Bebop Line: Jazz chromaticism, approach notes, and targeting chord tones.
7. Arpeggio Loop: A looped arpeggio cyclically moving through scale degrees.
8. Synth Arp: A dynamic synthetic line of 8ths and 16ths with a repetitive interval pattern (Synth-Pop / EDM).
9. Sequence Run: Ascending or descending runs of parallel thirds and sixths.
10. Ambient Drift: Wide interval leaps, long sustained notes, and spacious pauses (Ambient / Soundscape).
11. Tension Build: Tension building by raising register and rhythmic density.
12. Modal Motive: Ethnic modal motifs and pentatonic gaps.

3. Progressions & Harmony (Piano Roll)

The PROGRESSION & HARMONY section of the piano roll's right panel makes a chord progression, stores it as the clip's chord lane and plays it as notes.

Sources (tabs):

• LIBRARY: 240 named progressions in 20 categories (pop, rock, jazz, blues, gospel, soul, lo-fi, EDM, cinematic, classical, modal, chromatic, Latin, funk, metal, J-Pop, hip-hop…), searchable by name, genre and mood. KEY FIT decides how a progression meets the editor's key: IN SCALE (a minor progression in a major key starts on the relative minor, Am–F–C–G in C major), SAME ROOT (same tonic, the progression's own mode), REMAP (every degree takes the editor key's chord). Borrowed and applied chords always stay themselves. The current lane can be saved to My progressions (.progomni files in <config>/progressions/).
• GENERATE: 18 styles (Pop, Rock, Ballad, R&B / Neo-Soul, Jazz, Blues, Gospel, Lo-fi, EDM, Cinematic, Classical, Bossa, Funk, Folk, Metal, J-Pop, Ambient, Dark). Knobs COLOUR (triads → sevenths → ninths → altered), TENSION, SURPRISE and BORROW (chords from outside the key), plus harmonic rhythm (1/bar, 2/bar, 1 per 2 bars, TURN, 2+1+1, 3+1, 1/beat), tension curve (FLAT/RISE/FALL/ARCH/WAVE), cadence (LOOP, V-I, HALF, IV-I, V-vi, OPEN), start on I, A A' form and a seed. Locked lane chords stay when it regenerates.
• MELODY: chords under the clip's melody (the selection, or every note the lane did not write), weighing the melody against the style.
• TYPE: chords as text: symbols (Am7 D9 Gmaj7), Roman numerals (ii7 V7 Imaj7, bVII, V7/vi), bar lines (| C | G | Am F |), . holds a chord, I:2 sets beats.
• DETECT reads the chords the clip's notes already spell; AI describes a progression in words (LLM provider).

SOUND shapes how the lane plays: voicing (Close, Drop 2, Drop 3, Drop 2+4, Open, Spread, Shell, Rootless A/B, Quartal, Guitar — playable fretboard shapes, Pad, Power), register, voice count, top-line rule (SMOOTH/FREE/HOLD/RISE/FALL), low interval limits; pattern (block, legato pad, 3-3-2 stabs, offbeat, pulses, arpeggios, strum, piano ballad, Alberti, Charleston, bossa, gospel push), anticipation (PUSH), velocity, gate, accent, strum, humanise; CHORDS / BASS / TOP parts and the bass (ROOT/SLASH/SMOOTH/PEDAL × sustain, eighths, octaves, root-fifth, walking, syncopated). Voices are led across the whole progression at once (least motion).

The chord lane above the grid: colour = function (tonic / predominant / dominant), symbol and Roman numeral. Click to hear a chord, drag its right edge to change its length, double-click to type a chord, right-click for reharmonisation (same-function substitutes, borrowed chords, applied dominants, ii-V into the next chord, tritone and backdoor subs, passing diminished, suspensions, neo-Riemannian moves, chromatic mediants, colour, inversions, pedal), chord quality, lock, split, merge and delete. With LINK on, lane edits rewrite only the notes the lane wrote; a melody drawn by hand stays. Every change is one undo step (notes and lane together). The panel's map shows the progression on the circle of fifths and its tension curve. The melody generator follows the lane's chords, and a progression dragged from the browser's audio analysis becomes the clip's lane.

4. Integration with LLM AI Models (AI MIDI Create & AI Progression)

For users looking for full generative synthesis, Omni DAW offers direct integration with LLM models. 6 providers are available: Anthropic Claude (default), OpenAI GPT, Google Gemini, DeepSeek, and two local ones — LM Studio and Ollama (no API key required, configured via local address, defaulting to http://localhost:1234/v1 for LM Studio and http://localhost:11434/v1 for Ollama):

• AI Create — Melody Window: Type a text prompt (e.g., "A dark bassline in synthwave style at 120 BPM in D minor") – the engine sends a query to the selected AI provider and inserts the generated notes onto the Piano Roll grid.
• AI Generate — Progression Window: Generating multi-chord harmony from verbal descriptions of the track's vibe.

5. AI Image Generator for the Iris Synthesizer

For the Iris synthesizer (see /docs/alchemical-synth — image-to-sound resynthesis engine), Omni DAW features a built-in AI image generator based on the Google Gemini model (gemini-2.5-flash-image, independent of the text model selected in AI settings):

• Type a visual description (e.g., "Aurora borealis over an arctic ocean"), and AI will generate an image that is automatically baked into the additive spectral table (frames × partials) driving Iris synthesis.

6. Audio to MIDI (Basic Pitch)

Omni transcribes the notes of any audio into MIDI with Basic Pitch, Spotify's open note-transcription network (Apache 2.0). The model is built into Omni and runs on your computer: nothing is uploaded and no download is needed. It hears several notes at once (chords, piano, guitar parts) as well as single lines (vocals, leads, bass), and three minutes of audio take about a second.

• Browser: hover a file in the Samples list and click the piano key (or right-click → Audio to MIDI). The AUDIO TO MIDI panel shows the notes over the file with the preview playhead. Drag the roll or DRAG MIDI onto a track, a clip or the piano roll; the notes sit on the file's tempo grid (from the file name or estimated, like the chord panel).
• Sampler: the piano key in the sample editor's toolbar draws the notes over the waveform. DRAG MIDI takes the notes of the playing region, on the sample's own tempo in Warp and the song's otherwise, shifted by the sampler's pitch. In Slice mode SLICE puts a slice at every note.

Settings (the analysis runs once; the knobs re-read it instantly): POLY (every note), MELODY (one line from A1 up) or BASS (one line up to G2); SENS (how readily a note starts), SUST (how long notes ring on), MIN (shortest note kept, in ms). For a mixed song, split it into stems first and transcribe the stem you want.