Introduction
Anyone who has attended my talks or workshops has certainly heard me say that the audiovisual industry will be the most impacted by generative AI models. generative artificial intelligence is beginning to play a significant role in television content creation processes. In areas such as pre-production, post-production, and distribution, professionals' decisions are being influenced by AI's ability to optimize and innovate processes, minimizing legal risks related to copyright.
1. Synthetic Voices and Dubbing
AI-generated voices have shown initial utility as they become more natural. Synthetic voice dubbing is gaining traction in "low-risk" content, such as localizing news or sports clips for platforms like YouTube, or programming for FAST (Free Ad-Supported Streaming TV) channels. In these cases, speed is essential to expand audience reach for content that otherwise wouldn't be dubbed.
However, dubbing premium television content still faces challenges. AI-generated voices can exhibit imperfections compared to human voice actors. While some flaws can be corrected by adjusting pitch and inflection, the effort required to achieve acceptable quality may not be worth it compared to traditional recording. For now, AI dubbing can be useful for expanding reach and monetization, especially in less resource-intensive languages that typically don't receive dubbed versions.
In addition to dubbing, the use of voice clones for narration has emerged, always with the consent and compensation of the professional or their estate. One example is the voice clone of Al Michaels used to provide personalized highlights of NBC's Olympics coverage on Peacock.
2. Face Swap and Lip Sync
Deep learning models are highly effective at complex or subtle facial modifications. The most promising initial applications are lip-synching in voice acting and face swapping for effects such as de-aging.
AI lip-syncing tools, such as those offered by Flawless and LipDub AI MARZ's technology can synchronize an actor's lip and facial movements with the dubbed audio track. Major Hollywood studios are testing this technology to provide a more immersive experience for foreign audiences, making it appear as if the content was originally produced in their native language.
Face swapping can also be used for cosmetic touch-ups or to completely alter an actor's appearance, whether to age or rejuvenate them. These tools also open up possibilities for eliminating reshoots, allowing actors to rewrite lines of dialogue remotely.
3. AI Video Generation
Video production is advancing rapidly, and studios and filmmakers are showing interest in incorporating these models as production tools. However, there are still uncertainties about how to professionally integrate them into workflows and who would be qualified to operate them. With significant differences from traditional filmmaking, visual effects, or animation, issues such as photorealism, consistency, and control are key areas of concern.
While criticism suggests that text-to-video generation can be unpredictable, techniques like video-to-video are emerging, as seen in Runway's recent launch of Gen-3 Alpha. Major studios are exploring fine-tuning video models, training them with proprietary content for internal use. The partnership between Lionsgate and Runway is a public example of this initiative, with other Hollywood studios following suit.
Conclusion
While the performance of generative AI continues to improve to meet premium television standards, pressing legal issues still pose significant barriers to its full adoption in content production. However, the opportunities these technologies present signal a promising transformation in the television industry, boosting creativity and efficiency in the production and distribution processes.
