- The open-ended model will empower India’s AI ecosystem.
AI Videos: The race for Artificial Intelligence (AI) is increasing across the world and now India is also rapidly making its mark in this field. Taking a big step in this direction, Avataar AI has launched a new video generation model named Varya. This model not only makes the video making process faster and cheaper but is also designed to better understand the Indian culture and context.
based on global technology
The basis of Varya is the open-source video generation model Wan 2.2 developed by Alibaba. However, instead of using it directly, Avataar AI has transformed its capabilities into a lighter and faster version. This process is called model distillation, in which a large AI model is made capable of working with fewer resources while maintaining its properties.
The company claims that with the help of this technology, Varya can produce videos at a much faster speed than before. While Wan 2.2 requires 50 steps to create a video, Varya does the job in just 4 steps. Due to this, the speed of video generation increases by almost 10 times and the cost also reduces significantly.
The video will be ready in a few seconds
According to Avataar AI, Varya can produce a 5-second long video at 720p resolution in approximately 45 seconds when using the NVIDIA H200 GPU. In comparison, Wan 2.2 takes about 1,230 seconds for the same task.
According to the company’s internal tests, Varya’s video generation cost is around 48 paise per second. This makes it far more affordable than many international video AI models.
Special training to understand Indian culture better
The biggest specialty of Varya is considered to be its cultural understanding. Avataar AI has trained it on specially selected data sets so that it can better recognize Indian lifestyle and cultural elements.
This includes the ability to understand things like a glimpse of festivals, traditional and regional costumes, Indian cuisine and the diverse architecture of the country. This will make videos created by AI appear more realistic and relevant to local contexts.
How can Varya be used?
Emphasis has also been laid on making this AI model easy to use. Users can simply write their idea using text or upload a photo. After this Varya prepares the video clip based on that input.
If users want to extend the video, they can also extend the same story or scene by generating additional clips.
Technical information will also be made public
Avataar AI has said that it will soon release a detailed technical report on Varya’s technical architecture, model distillation process and performance related benchmarks. This will help developers and researchers to understand the system of the model.
India’s AI ecosystem will get a new boost
Varya is not only being seen as an AI product but it is also being developed as an open ecosystem project. The company plans to make it available as an open-weight model through the AI Kosh platform.
This means that developers will be able to host this model on their servers, make changes in it as per the need and can also develop new AI applications based on it. This step can play an important role in further strengthening India’s rapidly growing AI ecosystem.
Also read:
You will be able to create and edit images even without internet, this fun feature has come in Canva
