Sponsored by
Hey Creator,
AI video has been moving fast, but for many creators it's still felt out of reach.
Every so often, though, a launch comes along that shifts the conversation from what's possible to what's practical.
This is one of those moments!
India Just Built a Video AI Model That Changes the Math for Creators
If you've ever opened a video AI tool, seen the price per second, and just... closed the tab — this one's for you.
Most AI video generators are priced for studios, not solo creators. $0.10+ per second adds up fast when you're testing five versions of a 10-second clip before you find the one that works. That math has kept a lot of creators on the sidelines, watching others use tools they can't justify the cost of.
Avataar, a Bengaluru-based startup, just launched a video model called Varya — and it's built specifically to break that math.
What Varya Actually Is
Varya is an indigenous video-generation AI model, developed under India's IndiaAI Mission and unveiled in New Delhi on June 12, 2026. It's a 14-billion-parameter model created by Avataar, the same team behind e-commerce video tools, backed by Peak XV Partners and Tiger Global.
It turns a text prompt or a reference image into a short video — and then lets you extend that video with more prompts to build out a longer story.
Why It's Different: The Price
This is the headline that matters for creators:
Varya costs around ₹0.48 ($0.005) per second of generated video
Compare that to Veo, Kling, Luma, or Runway, which typically charge $0.10+ per second
That's roughly a 20x price reduction
The trick behind the savings is a technique called distillation. Varya was built starting from Alibaba's Wan 2.2 model, then compressed down so it generates video in just 4 steps instead of the usual 50+. On an NVIDIA H200 GPU, that means a 5-second 720p clip renders in about 45 seconds, versus over 20 minutes for the original Wan 2.2.
Faster generation, lower cost, more room to experiment — that's the pitch.
But Does the Quality Hold Up?
This is the part creators actually need before switching tools, so here's the honest answer based on benchmark data comparing Varya 1.0 to Wan 2.2 at full steps.
The overall quality gap is small. Total score: 0.8106 for Varya vs. 0.8172 for Wan 2.2 — a difference of less than 1%. For a model running 4 steps instead of 50, that's a strong result.
Where Varya actually outperforms Wan 2.2:
Handling multiple objects in a scene (+0.0837 — the biggest improvement)
Motion and movement richness, aka "dynamic degree" (+0.0389)
Spatial relationships between objects (+0.0386)
Object recognition accuracy (+0.0249)
Human action accuracy (+0.0120)
Aesthetic quality and subject consistency (small gains)
Where it falls short:
Color accuracy is the one real weak spot, dropping noticeably compared to Wan 2.2
Scene accuracy also dips moderately
Background consistency, flicker, and imaging quality are slightly lower, though only by small margins
So: if your project leans heavily on precise color grading or brand-accurate palettes, double-check your output before publishing. For most everyday content — explainers, social clips, product demos — the gap won't be noticeable.
What Makes It Genuinely Useful for Indian Creators
Beyond price, Varya's other selling point is cultural accuracy. Most global video models are trained on Western-heavy datasets, so they routinely get Indian context wrong — clothing, festivals, food, architecture, everyday settings.
Avataar trained Varya specifically on Indian cultural data, so it's built to render:
Regional festivals and clothing correctly
Indian food, public spaces, and everyday scenes
Diverse community and regional representation
If you've ever generated a video with AI and had to fix or scrap it because it looked generically "Western," this is the direct answer to that problem.
Who This Is Actually For
Marketers and small businesses who need product or ad videos without studio budgets
Educators creating visual lesson content
Content creators who want to test more ideas without burning through credits
Anyone building for an Indian audience, where cultural accuracy isn't a nice-to-have
How to Try It
Varya is live now and open to anyone via Avataar's website using text prompts or reference images
It's also being released as an open-weight model on India's AIKosh portal (the government's public AI model repository), meaning developers can self-host or fine-tune it
Avataar has said it's open to partnerships with tools like Higgsfield and Adobe Firefly, so expect tighter integrations down the line
The Bottom Line
Varya isn't claiming to be the most powerful video model out there — it's not trying to beat Veo or Runway on raw capability. What it's doing is making AI video genuinely affordable and culturally accurate for creators who've been priced out or underserved by the current options.
If you've been holding off on AI video because the cost didn't make sense, this might be the model that finally does.
What’s next is almost here.
On July 16th at 1PM ET, beehiiv is going live with a look at the future of publishing, audience growth, and digital business.
What started as a newsletter platform has evolved into something much bigger: a place where creators and brands can grow, monetize, and own their audiences without stitching together half the internet to make it work.
The next chapter starts live at the Summer Release Event.
Join us to see what’s coming next.
