reloadium.com

generateVideoFromCarousel

Turn the carousel you JUST built with generateCarousel into a narrated video (MP4) — slides on screen, a voice reading over them, each slide shown for exactly as long as its own narration lasts. Use this when the user wants the carousel as a video / reel / presentation / voiceover rather than (or in addition to) still images. It always operates on the most recent generateCarousel result in this conversation, so call generateCarousel first (its format decides the video's shape: 4:5, 9:16 for Reels/Shorts/TikTok, 1:1 or 16:9); if there isn't one yet, this returns an error. Provide `narration`: an array with exactly ONE string per slide, in order, written to be SPOKEN in the SAME language as the slides' on-screen text (never the UI or browser language — the voice reads it verbatim, so French narration over English slides is a mismatch) — conversational, and short (roughly 15-40 words per slide; a slide is on screen only as long as its narration takes). Don't just read the on-slide headline back word for word; say what a presenter would say over it, and let the last one wrap up. If you omit `narration`, each slide's own headline and body are read aloud as-is (fine as a quick draft, flatter than a written script). Optional `voice` (a Gemini prebuilt voice name, e.g. "Kore", "Charon", "Zephyr", "Orus") and `style` (delivery direction such as 'upbeat and energetic' or 'calm, warm, unhurried') apply to every slide, so the whole video sounds like one presenter. This is slow and costs credits (one text-to-speech call per slide, run in parallel, then an in-browser video encode; expect roughly 20-60 seconds and a one-time download of the encoder the first time) — only call it when the user actually asked for a video, not as an automatic follow-up to every carousel.

Page: /

Input schema

{
  "type": "object",
  "properties": {
    "narration": {
      "type": "array",
      "items": {
        "type": "string"
      },
      "description": "Exactly one spoken script per slide, in slide order, in the same language as the slides' on-screen text. Omit to read each slide's own headline + body instead."
    },
    "voice": {
      "type": "string",
      "description": "Optional Gemini prebuilt voice name applied to every slide, e.g. Kore, Charon, Zephyr, Orus."
    },
    "style": {
      "type": "string",
      "description": "Optional delivery direction applied to every slide, e.g. 'upbeat and energetic'."
    },
    "format": {
      "type": "string",
      "enum": [
        "4:5",
        "9:16",
        "1:1",
        "16:9"
      ],
      "description": "Optional video shape; defaults to the carousel's own format. The video card also has format tabs the user can switch between, so only pass this when they asked for a specific shape. Calling again with just a different format reuses the existing narration (no new speech)."
    }
  }
}