OPENAI'S VIDEO GENERATOR, SORA, DISAPPOINTS; BIZARRE OUTPUTS AND SUBSCRIPTION FEE HURDLES STAND IN THE WAY OF 'REVOLUTIONIZING' VIDEO PRODUCTION
In the increasingly digital landscape of the 21st century, where anyone with a decent internet connection and basic know-how can, in theory, become their own video producer or content creator, OpenAI's new AI-driven video tool, Sora, promises to be a game-changer. Yet, based on our trials with its technology, it might not be ready for primetime just yet as the future waits for the generation of coherent, AI-generated video content.
Recently released by OpenAI, the same organization known for GPT-3, the world's most powerful language model, Sora is an AI video generator that can create videos from diverse prompts. The technology certainly sounds impressive – who wouldn't love to have their own AI video producer reporting for duty at the press of a button? The aim is streamlined efficiency – simply key in a description, and Sora could generate a new video based on your prompt.
However, the usage of Sora is not as straightforward as it promises to be. The system needs fairly simple descriptions to execute the command; anything too complex can confuse the AI, resulting in a jumble of visual inconsistencies and strange oddities in the final product. A paid subscription is also required to unlock Sora. Prices range from $20 a month for the "Plus" tier to an eye-watering $200 a month for the "Pro" version that offers all the features.
Our trial experiences with the Plus subscription option of Sora were, unfortunately, less than satisfactory. Despite the strikingly innovative technology behind this AI, it was underwhelming, specifically for advanced prompts. It stands out, nonetheless, that Sora surpasses its contemporary – Runway AI, another AI video generator, in delivering better photorealism.
Despite these promising technical advancements, none of the videos we generated from scratch were usable. The narratives lacked coherence, and various oddities, characteristic of AI-driven content, were evident, detracting from the user experience.
One of Sora's features is its built-in "guardrails" designed to prevent copyright infringement or improper content generation. Functional in theory, these guardrails showed varying levels of success during our testing.
Sora also allows users to upload their own reference images, which is a significant addition. However, to take full advantage of this feature, the soaring cost of the Pro subscription is necessary, potentially chilling enthusiasm for potential users.
For the everyday user or small-scale content producers, the cost concern combined with dubious quality output raises questions about the convenience-versus-cost balance of Sora's service in its current version. While it's a promising glimpse into a future where AI can shoulder much of the creative burden in content production, for now, Sora's democratization potential appears limited.
The future is undoubtedly going to be impacted by AI-generated content, but in the case of OpenAI's Sora, it seems that future still has a fair bit of buffering to do. It highlights the continued need for human creativity and the masterful touch only a human can put into a story, at least for now. For aspiring video producers and professionals in the field, rest assured: your jobs are secure, for Sora, while revolutionary, is still a long way off from replacing human creativity and imagination.