AI voice cloning with an authorized voice reference
A voice-cloning workflow uses a reference or voice setup to guide synthesized speech. Compare the available routes below and use only a voice you own or have permission to synthesize.
Confirm the purpose and the voice reference
Start with the intended script, audience and an authorized voice reference. Use a clear source recording and inspect the model requirements before uploading it. Reference-audio fields, text fields and supported settings can differ between the Qwen and MiniMax workflows below. Do not treat one route as a drop-in replacement for another or assume that their source limits are identical.
Cloning, voice design and text to speech
Voice cloning is different from describing a new voice in words or choosing a supplied voice for ordinary text to speech. Select a workflow that matches the material you actually have. If you only need a narrated script, the text-to-speech guide may be a better starting point. For a two-way business conversation, use the voice-agent setup guides instead of expecting a single generated audio file to act as an agent.
Review the model inputs and credits
Guests can inspect public inputs and prepare a reference selection. Verified sign-in is required before calculating the credit estimate or generating. Check the current model information, required inputs and available settings; review the estimate before dispatch. Keep the same model page open for saved results. This guide does not run a new cloning request or promise free generation.
Evaluate the generated speech
When you decide to generate, compare the spoken result with the intended script and the approved voice use. Listen for pronunciation, pacing, unintended wording and artifacts. A single output does not establish consistent similarity or quality across speakers, languages and recordings. No fresh similarity benchmark or customer outcome is claimed here.
Open the available workspaces
Continue with these resources
Common questions
Can I use any voice found online?
Use a voice you own or have permission to synthesize. Public availability of a recording is not presented here as permission to clone its speaker.
Do these workflows have identical inputs?
No. Open each available workspace and check the current required fields and supported settings.
Is a cloning output automatically a conversational agent?
No. A conversational agent also needs its own setup and deployment. A generated audio result is a different workflow.