Tone: whisper · Kore

Listen first, then inspect the transcript and the request that produced this file.

Listen to the demo

MalayOne voice0:20

As requestedgemini-3.8-flash-tts · Studio (prebuilt) · Style: whispered softly

Human-reviewedFemale voice · Fitting for this scene

Accent and pronunciation have not been verified by a panel of Malaysian speakers.

Exact transcript

Lampu bilik sudah dipadamkan, dan adik akhirnya tertidur. Perlahan-lahan, letakkan hadiah itu di atas meja kecil. Jangan gerakkan kerusi, nanti dia terjaga. Esok pagi, kita boleh lihat wajahnya apabila dia menemui kejutan ini.

Generation recipe

This is the request that produced the file. Voice identity and delivery style are separate. Other settings: Default (not sent).

“Copy for AI” (top of this page) copies everything in one Markdown block: the transcript, style, voice, model, the request and example code. “Copy recipe JSON” copies only the raw request.

Show request JSON

{
  "model": "gemini-3.8-flash-tts",
  "input": [
    {
      "type": "user_input",
      "content": [
        {
          "type": "text",
          "text": "Lampu bilik sudah dipadamkan, dan adik akhirnya tertidur. Perlahan-lahan, letakkan hadiah itu di atas meja kecil. Jangan gerakkan kerusi, nanti dia terjaga. Esok pagi, kita boleh lihat wajahnya apabila dia menemui kejutan ini.",
          "annotations": [
            {
              "type": "speech_metadata",
              "style": "whispered softly"
            }
          ]
        }
      ]
    }
  ],
  "response_format": {
    "type": "audio"
  },
  "generation_config": {
    "speech_config": [
      {
        "voice": "Kore"
      }
    ]
  }
}

Comparison without style

Same transcript, voice and model, with no style direction sent. Listen to both to hear what the style changes.