Recently, US-based startup Luma AI unveiled a new video generator called Dream Machine, similar to OpenAI's Sora, further blurring the thin line between the real and artificial worlds. Luma AI claims that the new tool is a next-generation video model that uses AI to create high-quality, realistic shots from natural language prompts.
“Dream Machine is an AI model that quickly creates high-quality, realistic videos from text and images,” reads the introduction on the official website. Dream Machine comes on the heels of Kuaishou Technology, a Chinese company known for its short video platform, announcing its text-to-video conversion model, Kling. The difference between Sora and Kling is that Dream Machine is publicly available.
Luma AI describes Dream Machine as a scalable, efficient Transformer model that can generate physically accurate shots, trained directly on video. The company says Dream Machine, a first step toward building a universal imagination engine, can generate 120 frames in 120 seconds and iterate faster. The tool reportedly generates five-second shots with smooth motion, cinematography, and drama. Dream Machine can also understand how humans, animals, and objects interact with the physical world, allowing you to create videos with character consistency and accurate physics, helping you experiment with an endless array of fluid, cinematic, and natural camera motions.
Flash forward to November 2022. Open AI's ChatGPT debuted, resulting in some surprisingly human-like works of art. One of those tools is Suno, an AI-powered music generator. Suno can create vocal and instrumental lyrics from simple prompts, directing you to an accurate genre. The free version of Suno is limited, but subscription-based plans allow you to use your songs commercially on YouTube and upload them to Spotify and Apple Music. Suno's FAQ section reads, “The availability and scope of copyright protection for content generated (in whole or in part) using artificial intelligence is a complex and dynamic area of law that is rapidly evolving and varies by country.”
Tools like Dream Machine and Suno have taken video and audio content generation to a new level, offering enormous commercial potential and the potential to dramatically reduce production time and costs. But the industry must address ethical and legal challenges around the use of training data and its potential for misuse to create deepfakes and spread misinformation. Despite these concerns, the allure of advanced, near-instant video and audio generation is likely to attract many creators and companies looking to push the boundaries of what's possible with AI-powered tools.
As the technology matures, it will be crucial for the companies involved to balance innovation, accessibility, and responsibility. Also lurking in the background is the broader debate about AI-generated content. The New York Times is suing OpenAI and Microsoft, alleging that ChatGPT was trained on millions of its articles without the company's permission. Does training an AI model on someone else's content constitute copyright infringement? This is a big, unanswered question.
The May 2023 hit “Heart on my Sleeve,” which was said to be made by Drake and The Weeknd and has been played more than 11 million times across various platforms, was revealed to have been created using AI by a user named Ghostwriter977. This resulted in a takedown notice from the artist's record label, Universal Music Group, and the copyright debate is still ongoing. However, policing AI-generated music and videos could pose unique challenges. The legal status of AI works remains unclear, tools to create AI works are widely available, and social media makes the distribution of AI works easier than ever.
