This ComfyUI tutorial workflow turns a plain-text script into a polished, multi-speaker dialogue using ElevenLabs v4. Each line you write becomes a distinct “turn,” and each turn can use a different voice—either a preset selected with ElevenLabsVoiceSelector or a custom voice created on-the-fly with ElevenLabsInstantVoiceClone. The core ElevenLabsTextToDialogue node sequences your turns in order and synthesizes a single, continuous audio file, so you end up with one mix-ready track for podcasts, game dialogue, or character conversations. Delivery cues like [quietly] or [laughs] can be embedded directly in your script to influence performance.

Two grouped paths are included: Option 1: Use Preset Voice and Option 2: Create a Voice Clone. If you want clones, bring in a clean reference using LoadAudio or capture one with RecordAudio, feed it to ElevenLabsInstantVoiceClone, and wire that clone to any turns you want it to speak. The template’s inputs parameter sets how many dialogue turns to render (1–10). It ships with 4 turns that alternate two voices by default. SaveAudioAdvanced writes the final result to disk, while Note and MarkdownNote provide inline guidance. Some nodes are bypassed when the template opens—select them and press Ctrl-B to enable before running.