Recently, there is a new product calledHotshot The new text-to-video AI generator is eye-catching, and the official video effects are quite amazing!
It is understood that this new Hotshot model can be up toGenerate videos up to 10 seconds long at 720p, and is very flexible in terms of creative style, and can show different animation styles, such as comic book-like animation or rotoscoping-style video. Its highlights can be summarized as follows:
- Text-to-video generation: Users can input text and generate 720p videos up to 10 seconds long, supporting a fast and convenient creation process.
- No watermark video output: Free user-generated videos have no watermark, which provides more freedom for creators.
- Diverse video styles: Hotshot can generate videos in a variety of styles, including comic styles and dynamic images, to meet different creative needs.
- High adaptability and scalability: The model has the potential to be extended to longer video durations and higher resolutions, and may also support audio generation in the future.
- Strong technical foundation: Hotshot is trained on 6 million video clips and uses a large number of GPUs to achieve efficient video generation.
The AI video generation model has now entered the public "early preview" stage, and users can experience this model for free on Hotshot's official website.However, it should be noted that the number of videos generated per day is limited to two, and the generated videos have no watermarks.
The team behind Hotshot completed model training in four months
The startup Hotshot was founded in 2023 by three founders: Aakash Sastry, John Mullan and Duncan Crawbuck.
"For the first time in the last decade, it's possible to build powerful and novel video applications for customers," Sastry said on Social Media X. This new Hotshot model is the foundation for them to achieve this goal, and they say it's just the beginning, with more surprises to come. Previously, Hotshot became popular as a free consumer AI photo creation and editing app, but now they are turning their focus to this new text-to-video AI model.
According to Sastry, the Hotshot team completed the training of this model in just four months, which is no small feat! They used 600 million video clips and relied on thousands of GPUs for training. The technical challenges in the process can be imagined, and sometimes even machine failures were a headache.
In order to ensure the smooth training of the model, their team members monitor the training process almost around the clock. Sastry also said that they are exploring longer video durations and higher resolutions, and may also add audio elements in the future.
In his view, AI-generated content may become mainstream in digital media within the next 12 months, especially in YouTube video production, where creators will be able to have full control over the generation process, from text to video and even audio.