For AI agents: the complete documentation index is available at https://docs.clickmax.io/en/llms.txt, the full documentation bundle is available at https://docs.clickmax.io/en/llms-full.txt, and this page is available as Markdown at https://docs.clickmax.io/en/features/max/train-ai-voice.md.
  • English
  • How do I train an AI voice?

    Under Max > Vozes (Voices), in the Clonar uma voz real? (Clone a real voice?) box, click Enviar gravação (Upload recording), name it, upload 1 to 3 recordings of the same person and confirm you have their authorization. The voice joins the library under the Minhas vozes (My voices) filter and works in any audio generated by Max, AI agents and the API.

    Before you start:

    • Role: training, deleting and changing the default voice need the permission to manage Voz com IA (AI voice). Listening and using voices only needs the permission to generate audio.
    • The Voz com IA feature must be on in the workspace modules, under the IA module.
    • Each workspace has up to 3 own voices, counting trained voices and voices created from a description. Delete one to train another.
    • Only train the voice of someone who authorized it in writing. Clickmax records who confirmed the authorization, when, and the text accepted.

    1. Prepare the recordings

    • One person speaking alone, with no music or other voices.
    • At least 1 minute in total. More time at the same quality makes the voice closer.
    • MP3, WAV, M4A, OGG or WEBM, up to 8 MB each.
    • Phone recordings work. Background noise is reduced during training.

    2. Train the voice

    1. Open Max > Vozes.
    2. In the Clonar uma voz real? box, click Enviar gravação. Training opens in three steps.
    3. Under Gravações (Recordings), drag the audio files onto the upload area or click Do computador (From computer). To use audio already in Drive, click Do Drive (From Drive), select 1 to 3 files and confirm. Play each recording to check it and watch the Áudio somado (Total audio) bar until it reaches 1 minute. Click Continuar.
    4. Under Identidade (Identity), fill in Nome da voz (Voice name) and, if you want, the Descrição (Description). A preview shows how the voice will look in the library. Click Continuar.
    5. Under Autorização (Authorization), review the summary, read the statement and check Tenho autorização da pessoa dona desta voz.
    6. Click Treinar voz.

    Audio uploaded through the Drive gallery is stored in Meu drive > Vozes, visible only to you. The Do Drive option shows up for people with Drive access.

    The voice shows up in the library within a few seconds, with the Do seu time (Your team's) badge. Click the card's play button to hear a sample sentence.

    3. Use the voice

    • To use it in every audio that does not pick a voice, select it in the Padrão do workspace (Workspace default) bar and click Salvar padrão.
    • For an AI agent to speak with it, click Usar no agente (Use in agent) on the card and pick the agent (How do I make the AI agent reply with audio on WhatsApp?).
    • Through the API or MCP, pass the voice id in the voice field of speech_generate. The id is in speech_voices_list, under workspace.

    Delete a voice

    On the voice card, click Apagar (Delete) and confirm. The voice is removed for the whole workspace, and whoever used it as default switches to the first Clickmax voice.

    If it did not work

    • The Enviar gravação button is disabled: the workspace already has 3 own voices. Delete one.
    • Recordings rejected: training goes back to the Gravações step. Swap in cleaner audio with only one person speaking, and try again.
    • File not accepted: check the format and size (up to 8 MB). This also applies to audio from Drive.
    • The Do Drive option doesn't show up: Drive is off in the workspace, or your role can't view Drive.
    • The Clonar uma voz real? box does not show: your role lacks the permission to manage Voz com IA.