Move over Netflix, there’s a new director in town! OpenAI has unleashed its latest AI creation – a virtual filmmaker called Sora that can generate stunning video footage from text prompts alone.
While other companies like Google and startups have revealed similar text-to-video projects, Sora stands out with its uncanny photorealism and ability to produce clips up to a minute long. Enough to craft full TV ads or social media scenes.
To showcase Sora’s skills, OpenAI shared several examples, including one depicting a bustling snowy day in Tokyo. The AI camera smoothly pans down a street, following pedestrians enjoying the weather against backdrops of falling sakura petals and snowflakes.
It’s an utterly convincing slice of Tokyo life – down to intricate details like passing cars, storefronts and even a masked passerby nod to COVID-19. Minute flaws like characters hitting a dead-end sidewalk reveal this is AI-generated. Yet the quality is lightyears beyond what many believed possible.
Prompt: A stylish woman walks down a Tokyo street filled with warm glowing neon and animated city signage. She wears a black leather jacket, a long red dress, and black boots, and carries a black purse. She wears sunglasses and red lipstick. She walks confidently and casually. The street is damp and reflective, creating a mirror effect of the colourful lights. Many pedestrians walk about.
In another demo, Sora brings a Pixar-esque monster to life using a simple text description – mastering fluid fur textures and expressions that once challenged animators at top studios like Disney.
But Sora’s secret weapon is an innate knack for cinematic storytelling and shot sequences. Without any human direction, the AI adopts panning angles and scene changes that unfold narratives in the generated videos.
“Sora did this automatically, learning about 3D geometry and consistency from seeing lots of data.” explains OpenAI researcher Tim Brooks.
This emergent filmmaking knowledge points to a technology that could empower amateur creators on TikTok and Instagram.
“The average person can make very high-quality content with Sora.” says Bill Peebles, another key researcher.
Of course, some film professionals may view Sora as a threat to human creatives. But Peebles believes new tools like Sora will enhance – not replace – the skills of directors. AI can’t mimic true artistry yet.
OpenAI plans to release Sora publicly once thorough safety testing is complete – perhaps later this year. As with DALL-E, strict content guidelines will prohibit harmful material in Sora-generated videos.
There are concerns though – misinformation and deepfakes being chief among them. The ability to animate a single image also worries some experts. OpenAI admits mitigating risks from tools like Sora requires industry-wide collaboration.
Copyright issues loom as another potential headache if Sora mimics protected film and TV scenes. OpenAI claims licensed and public domain footage makes up the training data. But legal disputes over what constitutes “fair use” seem inevitable.
Generation times under a minute point to efficiency, but the researchers won’t reveal exact speeds needed to create a minute of Sora video. Let’s just say you can grab a burrito while Sora works its magic!
Realistically, Sora marks only the beginning of AI-generated video media. Don’t expect the complete coherence needed to produce full feature films anytime soon. Continuity and control remain firmly human endeavours for now.
But we may not be far off from the first Oscar winning screenplay composed entirely of AI prompts! Sora’s arrival means the most creative filmmakers will view AI as the ultimate collaborative partner rather than a replacement.
Just as CGI forever changed Hollywood, creative evolution through tools like Sora seems certain. Just don’t write human directors out of the script yet. Their skills are safe, but the talent pool just got a whole lot deeper thanks to AI!
Discover more from TechBooky
Subscribe to get the latest posts sent to your email.