Inspiration
India's heritage is not just something we preserve. It is something people still practice, create, and pass down every day.
An artisan can spend days or weeks creating a product, but bringing that product into the digital world creates an entirely different workload. They need photographs, descriptions, pricing, translations, marketing content, social media assets, and a way to communicate the story behind their craft.
The problem is not that artisans lack stories. It is that the digital ecosystem often expects them to become photographers, copywriters, marketers, translators, and content creators as well.
We wanted to change that.
Our question was simple:
What if an artisan could start with what they already have, their craft, their photographs, and their own knowledge, and technology could handle the digital complexity around it?
That idea became Virasya.
We did not want to build another marketplace where traditional products become anonymous catalogue items. We wanted to build a system where the product, artisan, process, and heritage story remain connected throughout the digital journey.
What it does
Virasya transforms a small amount of authentic artisan input into a complete digital presence.
The artisan uploads real photographs of the product and can describe it using text or regional voice input. Our multimodal AI combines the visual information with the artisan's own description to structure the product into a detailed profile containing its title, category, materials, craft tradition, dimensions, description, and heritage story.
The system does not blindly fill gaps. If important information is missing, Virasya identifies it and asks the artisan to verify or provide it.
From this single verified product record, the platform can provide:
- AI assisted product structuring
- Browser based image enhancement
- Missing detail verification
- Market informed pricing assistance
- Multilingual product listings
- Instagram ready content
- WhatsApp ready messages
- Native script hashtags
- Heritage storytelling
- Buyer side product Q&A
- Procedurally generated heritage reels
Procedural Heritage Reel Engine
One of our most distinctive engineering components is the procedural heritage reel engine.
We did not want to build a button that simply places a product image inside a fixed video template.
Instead, Virasya takes the artisan's actual photographs, verified product information, and heritage story and dynamically constructs a vertical visual narrative.
Using Remotion, the engine renders 1080 × 1920 video at 30 FPS and adapts the composition through controlled movement, cropping, depth, transitions, and visual sequencing based on the available product content.
The product itself is never generated or redesigned.
The heritage story can also be converted into a voiceover that is synchronized with the video timeline. When the reel is played, paused, or scrubbed, the audio follows the corresponding position in the video.
This turns the same structured information used for the marketplace into a complete piece of social media content without requiring the artisan to learn video editing.
How we built it
Virasya is built using:
- Next.js, React, TypeScript for the application
- Firebase for authentication and cloud data
- Gemini for multimodal AI workflows
- WebAssembly and ONNX for browser based image segmentation
- Remotion for procedural video composition
- Web Audio API for timeline synchronized voiceover
The architecture revolves around a single structured product record.
Instead of building every feature independently, information captured once flows through the entire system:
Product Input → AI Structuring → Verification → Pricing → Translation → Marketing → Reel Generation → Marketplace → Buyer Experience
This makes the system more than a collection of AI features. Each stage builds on the verified information produced by the previous stage.
For image enhancement, we use browser based segmentation through WebAssembly and ONNX. The original photograph is processed in the browser and the actual craft remains unchanged.
For multilingual support, the verified English information acts as the canonical reference. Regional translations are generated from this source, allowing updates to propagate without silently retaining outdated versions.
For the reel engine, Remotion provides frame based composition and rendering. The system combines real uploaded assets with structured product information to create a dynamic visual sequence rather than relying on a single fixed slideshow.
For audio synchronization, the voiceover is mapped to the video timeline so that playback state and seeking remain synchronized.
Challenges we ran into
The hardest challenge was not getting an AI model to generate text.
It was deciding where AI should stop.
For a heritage platform, hallucinating a material, origin, traditional technique, or historical detail is not a harmless mistake. It can change the identity of the craft.
We therefore designed Virasya around a strict principle:
AI assists the artisan. It does not become the artisan.
Missing information is surfaced for confirmation instead of being silently invented. Product photographs remain authentic instead of being replaced by generated imagery.
Another major challenge was the reel engine.
A conventional template based approach would have been much easier, but it would produce repetitive content and would not scale elegantly across products with different numbers of photographs, stories, and visual characteristics.
We therefore built the reel generation as a procedural system that can construct different compositions from the available content.
Synchronizing generated video with voiceover created another engineering challenge. The audio could not simply play independently because users need to pause and seek through the generated reel. We therefore tied audio playback to the video timeline.
Finally, connecting AI, image processing, translation, pricing, marketing, video generation, audio, and marketplace functionality into one coherent workflow required us to think about the system as a product rather than a collection of demos.
Accomplishments that we're proud of
We are proud that Virasya became a working end to end system rather than a collection of isolated prototypes.
A single artisan input can flow through:
Capture → Understand → Verify → Price → Translate → Market → Create → Sell
We are especially proud of the procedural reel engine because it demonstrates a different approach to AI powered content creation.
Instead of generating artificial representations of a product, we use the real product as the visual source and generate the presentation around it.
We are also proud of building authenticity into the architecture itself.
The artisan remains the source of truth. AI helps structure, enhance, translate, and communicate that information, while important uncertainties are surfaced for human verification.
Most importantly, we built the experience around the artisan's actual journey rather than around the technologies we wanted to showcase.
What we learned
We learned that building an AI application is fundamentally a systems problem.
A powerful model alone does not create a useful product. The difficult part is designing how AI interacts with human input, structured data, validation, media processing, and the final user experience.
We learned to treat AI output as assistance that requires context and validation rather than as unquestionable truth.
We also learned that seemingly simple features can involve significant engineering. Procedural video generation required us to think in terms of timelines, frames, composition, asset adaptation, and synchronized audio rather than simply generating a video file.
Most importantly, we learned that technology for cultural heritage needs a different definition of success.
Making something look more digital is not enough.
It has to remain authentic while becoming more accessible.
What's next for Virasya
Our next step is real world validation with artisans across different crafts, regions, languages, and levels of digital literacy.
We want to observe where artisans struggle, simplify the workflow further, and measure whether Virasya genuinely reduces the effort required to establish and maintain a digital presence.
Technically, we plan to expand regional language support, strengthen pricing intelligence with better market data, improve buyer discovery, expand the procedural reel engine across more craft categories, and develop richer artisan profiles and storytelling experiences.
We also want to explore stronger connections between artisans, regional craft traditions, geographical identities, and buyers looking for authentic products.
Our long term vision is not simply to put Indian handicrafts online.
It is to build a digital ecosystem where the people who preserve India's living heritage can participate in the digital economy without having to compromise the identity of their craft.
The craft stays theirs.
The story stays theirs.
Technology helps it travel further.
Log in or sign up for Devpost to join the conversation.