Voice Synthesis Models
Create natural-sounding speech with our AI voice models. Perfect for voiceovers, audiobooks, and content creation.
All Voice Models
Text-to-speech and voice cloning
MiniMax HD
Human-level speech synthesis
MiniMax HD produces remarkably human-like speech with natural intonation and emotion. Perfect for voiceovers, audiobooks, and professional audio content.
XTTS v2
Voice cloning model for multilingual speech
Learn how XTTS v2 handles consented voice cloning, multilingual speech, reference audio, quality limits, and when to use the online QuestStudio tool.
Choose a voice workflow
Start with the voice you need, then choose the model. For narration using a selectable synthetic voice, review MiniMax HD in Voice Lab. For a project based on your own voice or a reference you have permission to use, review the XTTS v2 reference workflow. These are different inputs; a script alone does not provide a voice reference.
Test a short passage containing the names, numbers and punctuation that matter in the final recording. Listen for pronunciation, pauses, pacing and distracting artifacts. Keep the passage consistent when comparing candidates, then revise the script or selected settings before generating a longer section.
Check the model's supported controls and generation estimate in the tool. A reference recording does not guarantee an exact voice match. The XTTS sample checker can help you review recording conditions, but its checklist does not analyze audio or certify output quality.