Paste the words you want spoken, choose or describe the delivery, review the live credit estimate, and generate. The page starts with the runnable Qwen3 TTS model rather than a placeholder.

Qwen3 TTS is the current default because it is live and priced in Voor. The model selector remains available so the page can support additional runnable text-to-speech models later.

Yes. Write the final script in the intended language and use the language controls exposed by the selected model. Ask a fluent speaker to check pronunciation, meaning, names, and locale.

Write short sentences, use paragraph breaks, spell unusual names phonetically when needed, and separate characters or major mood changes into different takes.

Yes. Give one coherent direction such as calm documentary narration or upbeat product host, then specify pace, energy, audience, and pronunciation priorities.

Reference-led voice features depend on the selected model form. Upload only your own voice or audio whose speaker gave informed permission for this specific use.

The generator uses the selected live model and submitted inputs to display a credit estimate. Review it before generating, particularly for long scripts.

Check your Voor plan, the selected model terms, script rights, and voice consent. Disclose synthetic speech when a platform, contract, or audience context requires it.

Listen on headphones and a phone speaker for pronunciation, skipped words, harsh consonants, breath artifacts, emotional consistency, timing, and appropriate loudness.