Varya

Posted on Last modified on Comments
sfg-2026
ForumIAS LATEST
  1. 09 July | Make Your UPSC Answers More Impactful with Adjectives by Ayush Sinha | Click Here to Watch →
  2. 10 July | From 6 Attempts to AIR 53: Kiran's UPSC Success Journey | Click Here to Watch →
  3. 11 July | Your Friends Reflect Your Values by Ayush Sinha | Click Here to Watch →

News: Avataar has launched Varya, a distilled video model built to make frontier video AI affordable, accessible and relevant for India’s next generation of users.

About Varya

Varya
Source: NDTV
  • Varya is a distilled video model built to make frontier video AI affordable, accessible and relevant for India’s next generation of users.
  • Launched by: It has been launched by Avataar, an AI-native transformation company.
  • Features:
    • It uses a distillation technique that reduces video generation from 50 steps to 4 steps, while maintaining comparable output quality. 
    • It can generate video at ₹0.48 per second, making it up to 10x more cost-efficient than several leading global video models.
    • It has been built to understand and generate culturally rich visual outputs across India’s regions, festivals, communities, food, clothing, public spaces and everyday life. 
  • How it works:
    • Users can type an idea, upload an image, generate a video, and continue the story through additional clips.
    • One prompt can become a lesson, an ad, a guide, a film or a memory. 

About Avataar

  • Avataar is an AI-native transformation company building domain-specialized AI products that help enterprises drive efficiency, unlock new operating models and build defensible IP-led capabilities. 
  • The company is focused on applying AI to real-world business and consumer use cases across high-growth markets.
  • Avataar was among the companies selected under the IndiaAI Mission to develop indigenous AI capabilities using subsidized national computing infrastructure.

About Distilled Video Generation

  • Distilled video generation is a machine learning model-compression technique in which a smaller, faster student model learns to replicate the outputs of a larger, more computationally intensive teacher model.
  • The process transfers the teacher model’s capabilities to the student while reducing redundant computations and improving efficiency.
  • Traditional video generation models often require 50 or more iterative denoising steps to transform noisy inputs into high-quality video outputs.
  • Through distillation, the student model learns to approximate the same results using only a few denoising iterations, significantly accelerating the generation process.
  • This approach enables faster inference, lower computational costs, and comparable output quality to the original teacher model.
Print Friendly and PDF
Blog
Academy
Community