YouDub WebUI: Open-Source AI Video Dubbing and Localization Powered by VoxCPM2

Introducing YouDub WebUI, an open-source local video localization tool integrating OpenBMB's VoxCPM2 for vocal separation, translation, zero-shot voice cloning,

tau · September 12, 2026

#YouDubWebUI #VoxCPM2 #OpenBMB #VideoLocalization #AIDubbing #OpenSource #VoiceCloning #DevTools

YouDub WebUI: Open-Source AI Video Dubbing and Localization Powered by VoxCPM2

Developer @liuzhao1225 has officially released YouDub WebUI, an open-source video localization and automated dubbing tool powered by OpenBMB's VoxCPM2 speech generation foundation model. Built as a local-first development tool, YouDub WebUI streamlines the entire cross-lingual content pipeline—from video downloading and audio track demuxing to subtitle translation, zero-shot voice cloning, and final video rendering—into a cohesive, near-one-click local workflow for global creators and developers.

YouDub WebUI open-source video AI dubbing and localization interface powered by OpenBMB VoxCPM2

Image source: https://github.com/liuzhao1225/YouDub-webui

Adapting video content across international borders has traditionally required stitching together fragmented tools for speech-to-text, neural machine translation, voice synthesis, audio stems mixing, and subtitle burning. YouDub WebUI bridges these isolated steps into a single unified web interface, enabling creators to preserve original speaker nuances while significantly reducing production overhead in an offline, self-hosted environment.

End-to-End Pipeline: From URL Ingestion to Final Dubbed Video

YouDub WebUI orchestrates multiple specialized AI models and media processing utilities into a continuous, automated pipeline:

  1. Video Ingestion and Audio Stem Separation: Users provide a video URL from YouTube or Bilibili. The system automatically fetches the source media and isolates the speaker's vocal track from background music (BGM) and ambient sound effects with high precision.
  2. Speech Recognition (ASR) and Translation: The isolated dialogue is transcribed into timestamped text and translated into the target language, maintaining synchronization with the cadence of the original speaker.
  3. Zero-Shot Voice Cloning with VoxCPM2: Utilizing OpenBMB's VoxCPM2 voice generation model, the tool performs zero-shot voice synthesis from short reference audio segments, faithfully replicating the speaker's original timbre, pitch, and emotional cadence in the new language.
  4. BGM-Preserving Audio Mixing: The newly synthesized speech track is dynamically balanced and mixed back with the preserved instrumental and environmental audio tracks, retaining the immersive atmosphere of the original footage.
  5. Subtitle Burning and Video Assembly: Localized subtitles are composited onto the video stream, producing a production-ready, fully dubbed video file ready for distribution.

Battle-Tested Production Stability Across 20,000 Videos

Unlike experimental proof-of-concept projects, YouDub WebUI arrives with demonstrated reliability across high-volume production deployments.

  • Proven on an 800,000-Follower Bilibili Channel: The software has served as the core production backbone for '黑纹白斑马' (Black-Striped White Zebra), a prominent Bilibili channel with over 800,000 subscribers. It has successfully powered the production of more than 20,000 full-length localized videos, validating its algorithmic resilience, queue management, and system stability under rigorous real-world workloads.
  • Bi-Directional Cross-Border Workflows: YouDub WebUI supports both English-to-Chinese dubbing for YouTube content and Chinese-to-English localization for Bilibili videos. This bi-directional capability provides practical infrastructure for international creators looking to bridge Western and Asian video platforms.

Local-First Architecture, Authentication, and Hardware Constraints

Designed from the ground up as a self-hosted, local-first tool, YouDub WebUI ensures complete data privacy and flexibility for customization.

  • ModelScope Weight Management: Model checkpoints and weights are managed through the ModelScope ecosystem, allowing developers to inspect, fine-tune, or swap individual component models to fit specific production needs.
  • Mandatory Argon2id Authentication: To protect web deployments across local networks or remote instances, the interface enforces cryptographic authentication using Argon2id password hashing. Administrators must configure a valid Argon2id hash in the YOUDUB_AUTH_PASSWORD_HASH environment variable before the WebUI can be accessed.
  • High-Performance GPU Requirements: Running VoxCPM2 alongside real-time audio separation and video encoding locally demands substantial GPU compute and VRAM. Workstation-grade hardware is recommended to prevent processing bottlenecks during intensive batch operations.

Sources