Idea generation
Produce a batch of starting points conditioned on your own production character instead of staring at an empty session.
For producers
A personalized model inside a production workflow
Producers do not need another song vending machine. What is genuinely useful is cheap, relevant options during the stage where nothing is committed yet — and a model that knows your catalog rather than the internet's.
Placement
Where it sits in a session
Catalog / stemsSoundSketch optionsChooseBuild in your DAW
ANYANO is upstream of your DAW, not inside it. You use it to decide what to build, then you build it properly. Nothing here mixes, masters or replaces arrangement work.
What it is good for
Practical workflows
Arrangement exploration
Hear the same idea with different energy and structure before committing to an arrangement.
Song-section generation
Target a specific section — an intro, a bridge, a drop — using the structure and tempo controls.
Creative variations
Branch a fragment that works into several related versions and choose with fresh ears.
Alternative directions
Take a finished-feeling demo somewhere deliberately different to test whether the original was the right call.
Breaking creative block
Generate around the stalled material rather than restarting from nothing.
Catalog exploration
Train on an archive of unfinished work and mine it for ideas you abandoned too early.
Pre-production
Make arrangement decisions before booking players or studio time.
Experimentation
Train element-focused Sounds from separated stems and see what a drum-weighted or bed-weighted model proposes.
Dataset craft
Getting a Sound worth using
- 01
Curate, do not dump
A focused dataset of representative material beats everything you have ever bounced. Stylistic coherence in the input is what makes output usable.
- 02
Use stem separation deliberately
Separate tracks and train on isolated elements when you want the Sound weighted toward one part of your production rather than the full mix.
- 03
Let preparation do its job
Normalisation and trimming exist because catalogs are inconsistent. Prepared datasets produce more predictable runs than raw file dumps.
- 04
Version instead of overwriting
Each run creates a new Sound version with its own stats, so you can compare a tighter dataset against a broader one rather than guessing.
Limits
What it will not do
It will not mix or master, deliver a labelled multitrack session, clone a vocalist on demand, or reliably hit a precise reference. It has no taste and no context about the project. Treat it as a fast, well-informed sketching partner whose suggestions you throw away most of the time.
FAQ
Frequently asked questions
Related
Keep reading
Sketch faster, decide better.
Train a Sound on your production catalog and use it upstream of the session.