How much audio do I need to clone a voice?

MiniMax’s API asks for 10 seconds to 5 minutes. Vocloner recommends 30 to 45 seconds of a single speaker in a quiet room. In both cases you need rights to the voice.

How is emotion directed in each?

MiniMax uses a per-request emotion parameter and parenthesized interjections on some models. Vocloner uses bracketed tags inside the text that you can place sentence by sentence. Test the same script in both.

MiniMax Audio alternative: Vocloner vs. MiniMax

MiniMax offers speech synthesis and voice cloning through MiniMax Audio and its developer API. Vocloner focuses on a web Studio for cloning an authorized voice and directing emotion with tags in the script.

Features and workflow

Scroll the table to see both platforms. With a keyboard, focus the table and use the arrow keys.

Feature comparison between Vocloner and MiniMax Audio.
Topic	Vocloner	MiniMax Audio
Cloning sample	Upload it in Studio. 30 to 45 seconds of clean, continuous speech is recommended.	The API guide accepts mp3, m4a, or wav from 10 seconds to 5 minutes, up to 20 MB.[2]
Emotion control	Tags inside the text, such as [excited], [whispering], [sad], or [laughing].	The API accepts an emotion parameter (for example happy, sad, calm, or whisper) and, on speech-2.8 models, interjections such as (laughs) or (sighs).[3]
Voices and languages	Supports Spanish, English, Portuguese, French, German, Japanese, and other languages.	Its text-to-speech page advertises 300+ voices across 32 languages.[1]
Workflow	Web Studio with private voices by default and dialogue with up to ten voices on paid plans.	Besides the web app, it documents an API with synchronous and WebSocket synthesis.[3][4]
Pricing	Plans with monthly or daily limits on Pricing; commercial use starts with Basic.	The documentation publishes API prices per character and per cloned voice. Check it for the current cost.[5]

Reviewed October 31, 2026 using the MiniMax Audio page and its API documentation. The requirements cited apply to the API. Prices are not copied, and quality or latency is not measured.

Questions to help you choose

How much audio do I need to clone a voice?: MiniMax’s API asks for 10 seconds to 5 minutes. Vocloner recommends 30 to 45 seconds of a single speaker in a quiet room. In both cases you need rights to the voice.
How is emotion directed in each?: MiniMax uses a per-request emotion parameter and parenthesized interjections on some models. Vocloner uses bracketed tags inside the text that you can place sentence by sentence. Test the same script in both.

Official sources

Checked October 31, 2026