It's no secret that we now live in a world where creativity—and even convenience—has reached new heights, thanks to Artificial Intelligence (AI). Ask almost anyone, and it's likely they’ve used some form of AI to assist with school or university assignments, work projects, or even daily tasks and life decisions. While it might encourage a bit of laziness, we have to admit that AI has become an invaluable tool, especially when we hit roadblocks and need quick solutions.
As AI continues to evolve, it's gone from writing and coding with precision to generating images. Now, the next big leap is upon us: AI-powered video generation, which is set to become the new standard for sourcing stock footage. This technology could redefine how we create and access visual content in the near future.
Meta introduces Movie Gen AI Video Generator
Meta has announced a groundbreaking advancement in generative AI with its new video generator, “Movie Gen,” which is capable of producing high-definition footage accompanied by sound. This unveiling follows closely on the heels of OpenAI's competing text-to-video model, Sora, although public access to Movie Gen remains unavailable at this time.
Movie Gen utilizes text inputs to automatically create videos, edit existing footage, or modify still images. The tool is capable of generating videos at 16 frames per second (fps) or 24 fps with resolutions of up to 1080p, integrating ambient sounds, sound effects, and background music. The initial footage is upscaled from a base resolution of 768 by 768 pixels, but early A/B testing indicates that Movie Gen surpasses both OpenAI's Sora and Runway's Gen3 models in video quality.
Meta has demonstrated Movie Gen's ability to generate personalized content from uploaded photos, as well as edit standard videos using text-based commands. For instance, one demonstration involved taking a still headshot of a woman and generating a video that depicted her enjoying a drink in a pumpkin patch. The tool can also make adjustments to existing videos, altering elements like style, transitions, and even the addition of entirely new features.
In one example, an innocuous video of an illustrated runner was transformed in various ways—showing him holding pompoms, changing the background to a desert, and even dressing him in a dinosaur costume—all accomplished through text prompts.
The underlying technology is driven by a robust 30-billion-parameter model, paired with a 13-billion-parameter audio model capable of generating up to 45 seconds of 48kHz audio content. While the current version lacks voice synchronization capabilities, this design choice reflects the ongoing development of the model.
Check Out How The Audio For Videos Turned Out:
Concerns and Implications
However, the training process for Movie Gen has raised concerns regarding ownership and the potential for misuse. Meta stated that the model was trained on a mix of licensed and publicly available datasets, comprising around 100 million videos, one billion images, and one million hours of audio. While the company has been somewhat transparent about the general scope of its training materials, the specific sources remain unclear, echoing earlier concerns regarding the use of data from Australian users for training its AI models.
As AI image and video generators continue to advance, apprehensions from creative professionals—such as filmmakers, photographers, and writers—about their livelihoods have intensified. This unease has been a focal point in recent labour disputes, including the historic joint Hollywood strikes by the Screen Actors Guild - American Federation of Television and Radio Artists (SAG-AFTRA) and the Writers Guild of America (WGA) last year.
Despite the impressive capabilities of Movie Gen, Meta emphasizes that its technology is not intended to replace human artists or animators. Instead, the company views these models as tools to augment creativity and offer new avenues for self-expression, especially for those lacking traditional artistic skills. Currently, Movie Gen is still in the research phase, allowing Meta and the broader AI community to explore its potential applications and implications before making it available to the public.
Imagine the possibilities once Meta releases Movie Gen to the public. It could even open up new income streams for individuals, especially if it proves capable of generating and editing stock photos and videos with precision. The ability to customize content with ease could be a game-changer for many creators.
However, the big question remains: will Movie Gen come with a price tag or subscription model? Given the power of the tool, it's likely that Meta will need to monetize it in some way.
Regardless, once Movie Gen is live, the creative potential will be limitless—but let’s also hope it isn’t misused for harmful purposes.
What are your thoughts about this? Let us know in the comments.
*Sources: Visual and Reference Credits to Social Media & various cross-references for context.
FOMO much? Follow our Facebook and Instagram for more updates.