Multi-character text to speech

Each line is synthesized in a Web Worker with a local speech model, then concatenated with short pauses. You can write a single block or a dialogue with different “characters” (language/voice presets). FlashKit does not host the WAV.

The first run downloads model weights. Voices are functional, not studio actors. For a product commercial, plan a human pass or a dedicated voice vendor.

How to Use

  1. Wait for the model to finish its first-time download if needed.
  2. Use single mode for one voice, or conversation mode to assign lines to characters.
  3. Keep lines short. Very long paragraphs increase memory use and monotone delivery.
  4. Generate, preview, and download the WAV. Normalize it in an editor if levels differ between voices.

Key Features

  • On-device synthesis after the model cache is warm
  • Single-voice or multi-line dialogue
  • WAV download — no FlashKit watermark
  • Language presets for common locales

Frequently Asked Questions

Does this call Gemini or a cloud TTS API?

No. Speech is generated by the browser worker. Instagram captions are the Gemini-backed tool, not this page.

Why is the first generate slow?

Weights download once. Later runs on the same browser are faster.

Can I pick a named celebrity voice?

No, and you should not. Presets are generic language voices, not likeness clones.

Related Tools