Skip to content

Generate music, sound effects and speech with Bali

Bali can create the sound of your game: background music, sound effects and spoken dialogue.

  1. Ask Bali, for example: “Create upbeat chiptune music for the first level” or “Make a short coin pickup sound.”
  2. Bali works out whether you need music or a sound effect and prepares a Generate Audio card.
  3. Adjust the prompt and the options below, then click Generate.
  4. Listen to the result and click Done. Audio is saved as a WAV file.

Watch: Generate a pickup sound effect (1:03, from the tutorial Building Collectibles & Scoring with Bali)

Option What it does
Model Choose from the available music models. The default is marked.
Duration (seconds) 1 to 60 seconds. The default is 15.
Instrumental Music without vocals. Shown for models that support it.
Seed (Optional) A number that makes results repeatable. Shown for models that support it.
Option What it does
Duration (seconds) 0.5 to 22 seconds. The default is 5.
Loop-friendly Makes an effect that repeats seamlessly, for things like engines, rain or ambience.
  • Describe the mood, genre and instruments. “Tense orchestral strings with a slow drumbeat for a boss fight” is better than “boss music”.
  • Describe the action and material for effects. “Heavy wooden door creaking open slowly” or “Short magical chime, bright and sparkly.”
  • Keep effects short. Most game sound effects work best under 2 seconds. Use Loop-friendly for anything that plays continuously.

Bali can voice your characters with text-to-speech, for dialogue, narration and announcer lines.

  1. Ask Bali, for example: “Voice Nova’s opening line: ‘We don’t have much time. Follow me.’”

  2. In the Generate Speech card, check the line, then set:

    Option What it does
    Model The speech model.
    Voice Name The voice to use for the character.
    Instruction Required. How the line should be performed, such as calm and warm, urgent whisper or slow and menacing.
  3. Click Generate, listen, and click Done.

Speech files are saved as WAV files in a folder for each character, under assets/tts/.

  • Use the same voice for a character every time so they sound consistent across scenes.
  • Direct the performance in Instruction, not in the line itself. Put emotion, pace and volume there.
  • Generate dialogue in batches. Ask Bali to voice a whole scene, and it prepares a request for each line.