From aff00df9547c1bacb901d214f893c86f43e3ffa3 Mon Sep 17 00:00:00 2001 From: Yasutake Yohei <61961825+yasutakeyohei@users.noreply.github.com> Date: Thu, 25 Jun 2026 14:04:05 +0900 Subject: Whisperページ: GPU選択の理由を追記(TPUは非対応) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/content/docs/whisper-to-ai-moji-okoshi.mdx | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src/content') diff --git a/src/content/docs/whisper-to-ai-moji-okoshi.mdx b/src/content/docs/whisper-to-ai-moji-okoshi.mdx index a0d44c6..646bcb7 100644 --- a/src/content/docs/whisper-to-ai-moji-okoshi.mdx +++ b/src/content/docs/whisper-to-ai-moji-okoshi.mdx @@ -75,6 +75,11 @@ description: OpenAI Whisperと生成AIを組み合わせた超高精度な文字 GPU設定画面 +:::tip +##### 💡 TPU ではなく GPU を選んでください +「v5e-1 TPU」などの選択肢もありますが、Whisper は PyTorch 製のため GPU(CUDA)でのみ高速動作します。TPU を選んでもエラーになるか、CPU 動作で非常に遅くなります。必ず **T4 GPU** を選択してください。 +::: + **⑥ 文字起こししたい音声ファイルを Google Drive の「マイドライブ」直下にアップロードします。** ここでは例として `onsei.mp3` で説明します。mp3、m4a、wav などの音声ファイルだけでなく、mp4 などの動画ファイルもそのまま文字起こし可能です。アップロードが完了するまで待ちます。 -- cgit v1.3.1