Skip to content

Sketch2Sound

Shape audio with text prompts and control signals

Visit official site

What Sketch2Sound is

The model is described as generating high-quality sound from text prompts together with time-varying control signals and sound imitation.

What you can do with it

  • Synthesise arbitrary sounds from sound imitation
  • Shape audio with loudness, brightness and pitch
  • Follow a text prompt while keeping audio quality

Getting started

  • Open the Sketch2Sound web page
  • Read the introduction to the functions
  • Watch the demonstration video
  • Supply a text prompt or sound imitation as input

Pros & cons

✓ What we like

  • Sound imitation can synthesise arbitrary sounds
  • Audio quality is kept while following input controls

! What to watch out for

  • Generation follows the input text prompt and controls
  • Control signals use median filtering during training

FAQ

Which controls are used?

The material names loudness, brightness and pitch as interpretable time-varying control signals.

How much fine-tuning is described?

The material states 40k steps and a separate linear layer for each control.

Last reviewed: 2026-11-15

More AI audio tools tools

View all →

How we review