Generation 1
Omni 1.0
Our first-generation Composer model, establishing the foundation for prompt-driven music generation.
- First-generation Composer model
- Prompt-driven music generation
- Efficient inference
- Foundation for subsequent Composer research
Composer
Music Intelligence for Composition
Composer is Falcore's music intelligence and generation model for creating instruments, arrangements, accompaniment, and complete musical material.
Give it musical intent, context, or a starting point. Composer helps turn that direction into structured musical material you can shape, revise, and use in production.
- Models
- Omni 1.0 · Omni 2.0
- Type
- Music intelligence and generation
What Composer does
Build the Musical Layer
Composer works across the parts of a composition that usually require different musical skills.
Instruments
Generate complementary instrumental parts around an existing idea.
Arrangements
Develop the structure, instrumentation, and movement of a piece.
Accompaniment
Create supporting musical material around a melody, vocal, or existing track.
Variations
Explore different musical directions without rebuilding the idea from scratch.
Complete musical material
Move from a brief or musical starting point toward a fuller composition.
Context
Music Should Understand the Music Around It
Composer is designed to work with musical context rather than treating every generation as an isolated request.
It can be used around existing material, creative direction, timing, instrumentation, and the intended role of the music in a larger production.
Workflow
From Direction to Usable Material
- 01
Define the intent
Start with a brief, musical idea, reference, or existing material.
- 02
Generate
Create instruments, arrangements, accompaniment, or broader musical material.
- 03
Explore
Compare variations and develop the direction that fits the work.
- 04
Shape
Keep control over the creative result and continue refining it for production.
Built for creators
Expand What You Can Compose
Composer is designed for creators at different levels of musical expertise.
Singers
A singer can build accompaniment without playing every instrument.
Producers
A producer can explore arrangements faster.
Filmmakers
A filmmaker can develop musical ideas around a scene.
Composers
An experienced composer can use it as another instrument in the workflow.
Production
Built for Production Workflows
Composer is being developed for real creative work, where musical material needs to fit timing, context, structure, and the rest of the production.
The goal is not just to generate music. It is to make musical intelligence useful inside the process of creating it.
The Composer models
Two Generations. One Direction.
Composer evolves through successive model generations, each focused on improving the quality and usefulness of prompt-driven music generation.
Latest generation
Omni 2.0
Our latest Composer model, developed with improvements across musical quality, prompt adherence and generation capability.
- Latest Composer generation
- Improved musical quality
- Improved prompt adherence
- Increased generation capability
Benchmark
Quality Is Measurable.
We evaluate Composer models across multiple dimensions rather than relying on a single measure of performance.
Omni 1.0 vs Omni 2.0 · Median of 9 runs per model
Omni 1.0Omni 2.0
Omni 2.071.6% lower
Omni 2.025.8% lower
Omni 2.013.1% lower
Omni 2.031.7% lower
Omni 2.013.1% higher
Omni 2.046.7% higher
Omni 2.096.4% lower
Omni 2.081.5% lower
| Metric | Better when | Omni 1.0 | Omni 2.0 | Omni 2.0 vs Omni 1.0 |
|---|---|---|---|---|
| Time to first event (s) | Lower | 4.90 | 1.39 | 71.6% better |
| Song saved, complete (s) | Lower | 78.71 | 58.42 | 25.8% better |
| Audio ready (s) | Lower | 105.42 | 91.65 | 13.1% better |
| Cost (Plumes) | Lower | 344 | 235 | 31.7% better |
| Length error (%) | Lower | 2.7 | 0.5 | 81.5% better |
| Length error, 30 s songs only (%) | Lower | 2.9 | 5.0 | Omni 1.0 leads |
| Loudness off target (LU) | Lower | 0.11 | 0.17 | 54.5% worse |
| Silence (%) | Lower | 1.86 | 1.91 | 2.7% worse |
| Notes in key (%) | Higher | 79.4 | 89.8 | 13.1% better |
| Semitone clashes (%) | Lower | 3.84 | 0.14 | 96.4% better |
| Bar variety (%) | Higher | 31.5 | 46.2 | 46.7% better |
| Distinct entry bars | Higher | 1 | 3 | 200% better |
| Parts sounding in bar 1 (%) | Lower | 100 | 50 | 50.0% better |
| Success rate (%) | Higher | 100 | 100 | No change |
Internal evaluation. 18 requests: three song lengths (30, 120 and 180 s), two models, three repeats, interleaved. Timings were taken on a single laptop, not production infrastructure, and each cell has three runs, so individual figures are noisy. Quality figures are computed from the MIDI and audio files, not from listening tests. Both models completed every request.
Omni 2.0 is faster, cheaper and cleaner on most measures. Omni 1.0 stays slightly closer on loudness target and silence, and is more accurate on length for 30-second songs.
Applications
Built for Creative Workflows.
Composer can be integrated into applications and workflows that need programmable music generation. These are potential applications.
Film and video
Games
Advertising
Creative applications
Prototyping
Interactive media
Generative content systems
Research
API
Build with Composer.
Integrate prompt-to-music generation directly into your application or creative workflow through the Composer API. Whether you're prototyping a new creative tool or building a production system, the API gives you programmatic access to Composer models.
Public Access Is Coming Soon.
Composer is in development as we work toward broader public availability. API access is available by request for teams and individuals who want to build with it.
- Status
- In development
- API access
- By request
- Public access
- Coming soon