Generate sound assets with a prompt

Use the Sound Generator tool to create custom audio clips from scratch with a text prompt.

To generate audio from natural language prompts, follow these steps:

  1. To open the Generate New Audio Clip window, right-click an empty area in the Project window.

  2. Select Create > Audio > Generate Audio Clip.

  3. To choose a model, select Change > Text to Sound.

    Note

    Currently, only one AI model is available for sound generation. After you select it, it will remain the default model for future generations.

  4. In the Prompt field, describe the sound effect you want to generate, such as jungle ambiance or robotic beep.

  5. To exclude specific elements from the generated sound, enter keywords in the Negative Prompt field. For example, no echo.

    For more information on negative prompts, refer to Remove unwanted elements with negative prompts.

  6. Use the Duration slider to specify the length of the generated audio clip in seconds.

    The model’s reference duration is 10 seconds, which produces the best results.

  7. Use the Count slider to specify the number of variations of the audio clip to generate in a single request.

  8. To specify a custom seed to generate consistent results, enable Custom Seed and enter a seed number.

    For more information on custom seed, refer to Use custom seed to generate consistent sprites.

  9. Select Generate.

The generated audio clip appears in the Generations panel. Hover over the audio clip to play it and view details, such as the model used and prompt settings.

Note

To generate and assign assets directly, refer to Generate and assign assets with the Object Picker.

Note

Sound Generator stores the generated audio files in the /GeneratedAssets folder located at the root of your project. These assets remain in that folder until you remove them manually.

Additional resources