Exploring the Frontier of AI in Advertising: Generative Video
Welcome to the AI in Advertising. We delve into the exciting developments in generative video—a challenging but revolutionary frontier in the generative AI industry.
Why is Generative Video So Complex?
At its core, a video is a sequence of images shown in rapid succession. Typically, it runs at 24 frames per second, meaning 24 distinct images are needed for every second of video to create the illusion of motion.
Generative video, therefore, requires AI not only to produce individual frames based on a text prompt but also to ensure those frames progress fluidly to simulate movement. This feat involves immense computational power, which is why current models typically produce clips lasting only about 10 seconds. However, platforms like Sora have pushed the limit to around a minute.
As with most technological challenges, these constraints will eventually be overcome. It's not far-fetched to imagine entire feature films created through generative AI in the near future.
Will AI Change Video Production Forever?
Since the advent of cinema in the late 1800s, filmmaking has adhered to a familiar formula: "Lights, camera, action." Over time, visual effects evolved from hand-drawn animations to advanced computer graphics, bringing us countless blockbuster films.
AI is poised to disrupt this process entirely. It can eliminate the need for traditional components such as scripts, sets, actors, and even the cameras themselves. For AI, generating Brad Pitt on Tatooine is no different from rendering a fictional superhero in a metropolitan skyline.
The implications are particularly profound for advertising. Most ads operate with smaller budgets than Hollywood productions, making them ideal candidates for AI-generated content. Generative video empowers creative directors to realize ambitious ideas without the constraints of location costs, actor fees, or extensive CGI work. For clients, the potential to replace a multi-million-dollar production with a few minutes of computational rendering is undeniably attractive.
The Current State of Generative Video Platforms
This year, several companies have made significant strides in generative video technology. OpenAI's Sora, Google's Lumiere, Meta's Make-A-Video, and Runway's Gen-3 are leading the charge. While most of these tools remain unavailable to the general public, they have been tested by select creatives.
Early results, such as a ToyRUs ad produced with Sora, have been mixed. While the quality and creativity are promising, the tools are still prone to errors. Publicly available options often produce inconsistent results, leaving users unable to fine-tune outputs effectively. However, it's crucial to remember that these are first-generation technologies. The potential for improvement is vast, and the future holds immense promise.
Challenges Facing Generative Video
Skeptics raise three main concerns about generative video tools:
Creativity Limitations: Critics argue that AI cannot be genuinely creative since it relies on pattern recognition and existing data. Yet, much of what we deem "creative" in films and ads builds on prior work.
Training Data Scarcity: Some fear AI development will plateau due to limited training data. Legal battles over copyright may restrict companies from using vast repositories of films and online videos. However, giants like Google are unlikely to face significant barriers in accessing such data.
Authenticity Concerns: The "uncanny valley" effect could deter consumers who value genuine, human-made content. However, as technology matures, this issue is likely to diminish. After all, if CGI can captivate audiences, AI-generated video can too.
The trajectory suggests we may see the first fully AI-generated feature film—from scriptwriting to production—within five years. While directors might still have roles to play, many other industry professionals may find themselves displaced.
Evaluating Current Generative Video Tools
Here is a snapshot of some of the leading platforms tested with a simple prompt and an image:
Runway
Time to Generate: About 5 minutes
Accuracy to Prompt: 6/10
Video Quality: 6/10
Cost: Approximately $0.10 per second of video. A free account offers 125 credits; $12/month provides 625 credits.
Verdict: The most sophisticated and production-ready platform, delivering impressive but sometimes eerie results. Achieving precise outputs remains a challenge.
Imagine.art
Time to Generate: Less than a minute
Accuracy to Prompt: 1/10
Video Quality: 2/10
Cost: Free with limited functionality; $8/month for standard features, $13/month for premium.
Verdict: Struggles with photo-realism and understanding prompts. Better suited for simple animations and transitions but far behind competitors like Runway and Sora.
Luma AI
Time to Generate: About 5 minutes
Accuracy to Prompt: 4/10
Video Quality: 4/10
Cost: Free for basic use; up to $400/month for premier features.
Verdict: Handles basic motion well but struggles with complex prompts. The quality declines over longer clips.
Pika Labs AI
Verdict: Geared towards cartoony animations and GIFs rather than serious video generation. Limited functionality for professional use.
Plazmapunk
Verdict: Specializes in music videos by analyzing tracks to generate abstract visuals. Current quality leaves professional producers unthreatened.
Honourable Mentions
While generative video tools are still evolving, AI is already making waves in post-production. Tools from Adobe, Apple, Avid, Da Vinci, and Synthesia incorporate AI for tasks like editing and special effects, significantly enhancing the current creative workflow.
Conclusion
Generative video is in its infancy, but its transformative potential is undeniable. Despite current limitations, the technology is advancing rapidly. It may soon reshape industries from advertising to filmmaking, freeing creativity from traditional constraints while raising critical questions about authenticity and human artistry
Validate your login
Sign In
Create New Account