MiniMax

MiniMax Voice Cloning

MiniMax Voice Cloning analyzes a reference audio file and returns a voice ID for compatible MiniMax speech models. Omnipix exposes a clone name, an accuracy control, and optional noise reduction and volume normalization. Clean, single-speaker source audio makes evaluation easier; use only recordings you are authorized to submit and reproduce.

How to use this model

  1. 1

    Choose a clean recording containing one speaker, with little background noise, music, or reverberation.

  2. 2

    Confirm that you are authorized to submit the recording and create speech in that voice.

  3. 3

    Give the clone a clear name and upload the reference audio file.

  4. 4

    Set accuracy deliberately; enable noise reduction or volume normalization only when the source needs it.

  5. 5

    Create the voice ID, then test it with a short, representative script in MiniMax Speech before a longer production.

Voice CloningAvailable

Voice Cloning

MiniMax Voice Cloning accepts a reference audio file and a name, then returns a voice for compatible MiniMax speech generation. Accuracy can be adjusted, and source cleanup can optionally include noise reduction and volume normalization. The best evaluation comes from testing the returned voice on a short script that differs from the reference recording.

Try now

Properties

  • Requires a name and an audio reference file.
  • Provides adjustable accuracy from 0 to 1.
  • Offers optional noise reduction and volume normalization for the source.
  • Returns a voice ID for compatible MiniMax speech generation.

Best for

  • Personal voiceovers
  • Consistent character voices

Tips

  • Use a clean recording with one speaker and minimal background noise.

How to use this model

  1. 1

    Confirm authorization to use the speaker's recording and synthesized voice.

  2. 2

    Upload a clean, single-speaker reference file and assign a recognizable name.

  3. 3

    Set accuracy, then enable noise reduction or volume normalization only when the source requires cleanup.

  4. 4

    Create the clone and note the returned voice ID.

  5. 5

    Test the voice ID in MiniMax Speech with a short representative script before longer use.

Parameters

Voice Name

name
Required

A private label for this voice in your library. It is not sent to MiniMax.

Voice Recording

voice_file
Required

Upload audio lasting 10 seconds to 5 minutes and under 20 MB. Unsupported audio formats are converted to compatible WAV.

Recording Acceptance

accuracy
Required

How strictly MiniMax validates the recording before cloning. Keep the recommended 0.7; lower it only if a clear recording is rejected.

Default: 0.7Range: 0 – 1

Noise Reduction

need_noise_reduction
Required

Turn this on when the recording contains background noise. Leave it off for a clean recording to avoid altering the voice.

Default: false

Volume Normalization

need_volume_normalization
Required

Turn this on when the recording is too quiet or its loudness varies noticeably. Leave it off when the volume is already even.

Default: false