Home
Alibaba's Tongyi Unveils Fun-CineForge: Open-Source AI Model Achieves Film-Grade Voice Synthesis
Alibaba Tongyi Lab officially launched and open-sourced the film-grade, multi-scenario voice synthesis multimodal model Fun-CineForge on March 16. This model addresses core challenges in AI dubbing, including lip-sync mismatch, lack of emotional expression, and inconsistent voice characteristics across multiple characters. It also introduces a high-quality method for dataset construction.

Technically, Fun-CineForge pioneers the concept of "temporal modality." Unlike conventional models that focus solely on text or visuals, it ensures voice synthesis occurs within precise time intervals through accurate timestamp control. Even in complex film scenes with occluded characters, frequent camera cuts, or blurred faces, the model maintains a high degree of audio-visual synchronization and adherence to instructions.
The accompanying open-source CineDub dataset construction pipeline is another key innovation. Tongyi Lab employed large language model chain-of-thought reasoning to automatically transform raw film footage into structured data, significantly reducing the need for manual annotation. This process achieves a word error rate of approximately 1% and a speaker diarization error rate of just 1.20%, providing a highly competitive training foundation for large models.

Fun-CineForge is now available on GitHub, HuggingFace, and the ModelScope community, supporting inference for video clips up to 30 seconds long. It excels not only in single-speaker monologues but also offers professional-grade support for duet and multi-speaker dialogue scenarios. This advancement signals AI voice technology's evolution from basic customer service and assistant roles into high-standard animation and film post-production.
GitHub: https://github.com/FunAudioLLM/FunCineForge
HuggingFace: https://huggingface.co/FunAudioLLM/Fun-CineForge
ModelScope: https://www.modelscope.cn/models/FunAudioLLM/Fun-CineForge/
Related article
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage
California AV Compliance: A New Era of Tickets, Geofences, and 1M Miles
Guident operates an AuveTech shuttle in South Florida, managing a four-mile route in West Palm Beach and a one-mile route in Boca Raton using its remote monitoring technology. | Credit: GuidentCalifornia is redefining the regulatory landscape for dri
Related Special Topic Recommendations
Comments (1)
0/500
Just tried the demo and honestly blown away by how natural the lip-sync feels now! 😮 Always thought AI dubbing sounded a bit robotic, but this seems like a huge leap. Wonder if this will start being used in indie films or even gaming soon? The open-source move is pretty bold too—curious to see how other companies respond.
Alibaba Tongyi Lab officially launched and open-sourced the film-grade, multi-scenario voice synthesis multimodal model Fun-CineForge on March 16. This model addresses core challenges in AI dubbing, including lip-sync mismatch, lack of emotional expression, and inconsistent voice characteristics across multiple characters. It also introduces a high-quality method for dataset construction.

Technically, Fun-CineForge pioneers the concept of "temporal modality." Unlike conventional models that focus solely on text or visuals, it ensures voice synthesis occurs within precise time intervals through accurate timestamp control. Even in complex film scenes with occluded characters, frequent camera cuts, or blurred faces, the model maintains a high degree of audio-visual synchronization and adherence to instructions.
The accompanying open-source CineDub dataset construction pipeline is another key innovation. Tongyi Lab employed large language model chain-of-thought reasoning to automatically transform raw film footage into structured data, significantly reducing the need for manual annotation. This process achieves a word error rate of approximately 1% and a speaker diarization error rate of just 1.20%, providing a highly competitive training foundation for large models.

Fun-CineForge is now available on GitHub, HuggingFace, and the ModelScope community, supporting inference for video clips up to 30 seconds long. It excels not only in single-speaker monologues but also offers professional-grade support for duet and multi-speaker dialogue scenarios. This advancement signals AI voice technology's evolution from basic customer service and assistant roles into high-standard animation and film post-production.
GitHub: https://github.com/FunAudioLLM/FunCineForge
HuggingFace: https://huggingface.co/FunAudioLLM/Fun-CineForge
ModelScope: https://www.modelscope.cn/models/FunAudioLLM/Fun-CineForge/
DeepMind CEO Hassabis: I sleep six hours a day, usually feel energetic around 1 a.m.
Fortune recently featured an interview with Demis Hassabis, CEO of Google DeepMind, revealing his unconventional approach to rest and productivity. Hassabis disclosed that he sleeps very little, structuring his waking hours into two distinct work blo
OpenAI, Anthropic Vie for Market Share Despite Revenue Shortfalls
Despite recent reports suggesting OpenAI missed revenue targets, creating pressure on tech stocks this Tuesday, private AI lab investors remain resilient. Seasoned backers have confirmed they will not reduce investment despite negative media coverage
Just tried the demo and honestly blown away by how natural the lip-sync feels now! 😮 Always thought AI dubbing sounded a bit robotic, but this seems like a huge leap. Wonder if this will start being used in indie films or even gaming soon? The open-source move is pretty bold too—curious to see how other companies respond.











