Short answer: there’s no documented script markup that makes ElevenLabs skip speaker names while also auto‑assigning voices. The TTS engine will read whatever text is in the prompt/import.
Supported workflows:
Use Dialogue mode (v3)
Enter each utterance as a separate speaker line and choose a voice per speaker; don’t include the names in the line text. This produces multi‑speaker audio without spoken labels, either on the website or via the Text‑to‑Dialogue API. (help.elevenlabs.io)
Use Studio for audiobooks and assign voices to text spans
Import your manuscript, delete speaker labels from the text you plan to synthesize, and assign voices to each character’s spans or paragraphs. You can assign multiple voices within a single paragraph or change voices across many paragraphs. (help.elevenlabs.io)
Notes:
If you keep “CHARACTER:” labels in the text, they will be spoken; the Help Center does not document a way to auto‑hide them in synthesis. (help.elevenlabs.io)
If you’d like, I can outline the exact steps for Dialogue mode or Studio based on your current workflow.
noble kettle
sick crystal