This ComfyUI workflow performs voice conversion using the ElevenLabs Speech-to-Speech API. You provide a source audio clip (via LoadAudio or RecordAudio), choose a target voice (via ElevenLabsVoiceSelector for a preset voice or ElevenLabsInstantVoiceClone to create a new clone), and the ElevenLabsSpeechToSpeech node generates a new recording that preserves the timing and content of the original while changing the vocal timbre to match the selected voice. The output can be saved with SaveAudioMP3 or SaveAudioOpus.

Technically, the pipeline routes your source waveform into ElevenLabsSpeechToSpeech along with a target voice ID. In Option 1, ElevenLabsVoiceSelector queries your ElevenLabs account and outputs a selected voice for direct use. In Option 2, ElevenLabsInstantVoiceClone uploads a clean reference sample and returns a new voice resource that is then fed into the Speech-to-Speech node. The Note node contains guidance, and several nodes are initially bypassed so you can enable only what you need. This design makes it easy to A/B between preset voices and instant clones while keeping file management simple with explicit save nodes.