- Stability AI unveils AI models, SVD and SVD-XT, generating high-quality videos from still images.
- The models have limitations, and Stability AI is transparent about training data, emphasizing future developments.
- Facing financial issues, Stability AI raises $25 million, setting ambitious goals despite executive departures and challenges.
Amidst the OpenAI leadership turmoil, Stability AI boldly announces "Stable Video Diffusion," an AI model capable of generating videos by animating existing images. The models, SVD and SVD-XT, are available in a research preview, with applications in "educational or creative tools" emphasized.
The release, though innovative, raises concerns about potential misuse and the absence of content filters.
Two models, limitations, and high-quality outputs
Stable Video Diffusion comprises two models, SVD and SVD-XT, capable of transforming still images into 576x1024 videos with 14 and 24 frames, respectively.
Despite generating high-quality four-second clips, the models have limitations, such as the inability to produce videos without motion or slow camera pans, control by text, or consistently render faces and people correctly.
Transparency and future developments
Stability AI remains transparent about the model's training data, primarily from public research datasets, potentially raising legal and ethical questions. The company acknowledges limitations but expresses the models' adaptability for use cases like generating 360-degree views of objects.
Stability AI plans to extend and commercialize Stable Video Diffusion, exploring applications in advertising, education, entertainment, and beyond.
Financial challenges and executive departure
Stability AI, facing financial challenges and executive departures, recently raised $25 million through a convertible note. Reports highlight issues with wage payments and payroll taxes, with AWS threatening to revoke access to GPU instances. The startup's valuation and fundraising goals are ambitious, adding complexity to its precarious financial situation.
Edited by Shruti Thapa
