A portable ComfyUI workflow for replacing dialogue within an exact video time range using OmniVoice voice cloning.

Features include:

• Synchronized video and waveform preview

• Millisecond-accurate start and end selection

• OmniVoice voice cloning

• Whisper transcription support

• UVR dialogue/background separation

• Stereo background preservation and automatic level matching

• Exact-duration speech fitting

• Iterative editing with saved-version history and rollback

• Lossless WAV/FLAC export for external lip-sync processing

The download includes the workflow, custom SpeechSwap nodes, an automatic Windows installer, setup documentation, dependency links, and model-download instructions. Model weights and source media are not included.

This workflow does not perform lip-sync internally. Exported video and audio can be processed afterward using an external lip-sync application.

Requirements: ComfyUI, OmniVoice-TTS, ComfyLiterals, FFmpeg, audio-separator, and the MDX23C InstVoc HQ model. An NVIDIA GPU is strongly recommended.

Only clone voices and modify media when you have permission. Users are responsible for following the licenses of all third-party models and software and for obtaining the necessary rights to voices, recordings, and source videos.