Harmony and melody
Recurring chord movement, intervals and melodic contours across the material you provided.
Product concept
A Sound is your model, trained on your music
Where other tools give you a prompt box in front of somebody else's model, ANYANO gives you a Sound: a private model built from your own catalog, versioned as it grows, and used as the starting point for everything you generate.
What it is
Definition
A Sound is a trained model derived from a dataset of material you uploaded, owned by your account, referenced by name, and available to the generator once a training run finishes.
Your MusicDatasetTraining runSound versionGeneration
Lifecycle
How a Sound is trained
- 01
Choose the material
Group files from your library into a dataset. What you include defines what the Sound leans toward, so a focused selection usually beats an exhaustive one.
- 02
Prepare it
Preparation normalises levels, trims dead space and computes the representations the run consumes, so inconsistent source material does not skew the result.
- 03
Configure the run
Pick what the run should focus on, set basic hyperparameters, and confirm your rights to the material — including any voice recordings.
- 04
Train and version
The run produces a new version of the Sound with its own record of songs used, time taken and size. Previous versions remain available for comparison.
Characteristics
What a Sound models
Rhythm and feel
Groove tendencies, typical tempo ranges and how your material tends to sit in time.
Timbre and production
The sonic character of your recordings — instrumentation, texture and production choices.
Arrangement habits
How your material typically moves between sections and builds or releases energy.
Element focus
Train on separated stems when you want a Sound weighted toward drums, an instrumental bed or a vocal approach.
Version drift
As you retrain with new material, versions let you hear how your Sound is moving over time.
Generation
Using a Sound to generate
In the generator you select a Sound, then shape a request around it: a prompt, optional lyrics, mood, structure, tempo and how far the model is allowed to stray. The Sound supplies the musical foundation; those controls steer within it.
Results arrive in your generations feed with a waveform player. From there you can branch into variations or extend a section — the loop described on the Directions page.
Control
Managing and deleting Sounds
Your Sounds are listed with their versions and training stats so you can tell them apart months later. You can delete a Sound or a version you trained, and delete uploaded files from your library independently.
Sounds are private to your account. They are not published, browsable by other artists, or used to train anybody else's model. Where third-party compute is involved in training or generation, that is disclosed on the privacy page rather than hidden.
Before training
Rights requirements
You may only train a Sound on material you own or otherwise hold sufficient rights to use for this purpose. That includes underlying compositions, recordings, performances, samples and any co-writer or featured-artist interests. Voice recordings deserve extra care and get their own confirmation step.
ANYANO is not a tool for imitating other artists. If you do not have rights to the material, do not upload it.
FAQ
Frequently asked questions
Related
Keep reading
Give your catalog a second life.
Train a Sound from music you already own and start exploring.