Text to speech and transcripts
VoiceLabs is a browser-based voice studio for text-to-speech and transcription. The VoiceLabs plugin lets ChatGPT work with your voice library directly: list the voice profiles on your account, provision a built-in preset voice when one is missing, generate spoken audio from text, transcribe an audio clip to text, browse your saved captures, and check the status of a generation you started earlier.
Access is granted through OAuth 2.1 with a consent screen that shows exactly which scopes you are approving: voice:read for browsing your library, voice:generate for producing audio and transcripts. You can revoke a connection at any time from Connected Apps in the VoiceLabs app, which invalidates both the stored consent and any live access tokens. Voice cloning is deliberately excluded from this surface. The plugin can only provision VoiceLabs' own preset voices, because uploading samples of a real person's voice requires a consent an agent cannot hold on your behalf.
Generated audio is returned as a short-lived signed link rather than a permanent public URL, and every tool call is recorded on the account's audit trail with its outcome.
VoiceLabs に接続すると、ChatGPT はリクエストのコンテキスト提供に役立てるため、関連するチャットやメモリをこのアプリと共有する場合があります。VoiceLabs によるこのデータの使用には、規約およびプライバシーポリシーが適用されます。メモリが有効な場合、アプリのデータは、役立つ情報や提案を先回りして提供するために使用されることがあります。ChatGPTは、連携アプリのデータを含め、トレーニングデータに関する設定を常に尊重します。アプリの使用には高いリスクを伴う場合があります。設定から、いつでも各種設定の管理やアプリの接続解除が可能です。詳細を見る