DEV Community: Swarnabha Mitra

From Chaos to Cinematic: A Bengali Durga Puja, a Small AI Startup, and a Big Dream ❤️

Swarnabha Mitra — Sun, 11 Oct 2026 03:24:18 +0000

Not every startup begins with thousands of users. Sometimes, it begins with a festival, a little hope, and a few people who believe in what you're building.

If you're Bengali, you know that Durga Puja is not just another festival.

It's an emotion. ❤️

The sound of dhak, the beautiful pandals, the glowing streets, the excitement of meeting old friends, and those unforgettable moments with family.

For five beautiful days, everything feels different.

And every year, we try to capture those moments through photographs and videos. But sometimes, a simple photograph cannot express the emotion behind a memory.

That's where an idea started growing in my mind.

🎬 What If We Could Turn Memories Into Cinematic Stories?

I'm an independent CGI artist and graphics designer from India. My background is mainly in Unreal Engine, Blender, VFX, and digital content creation.

A few months ago, I started experimenting with something outside my usual workflow.

Could I build a platform where ordinary people could create AI-generated videos without spending a fortune?

Not another complicated professional tool.

Something simple, accessible, and affordable.

That experiment eventually became "YourFXCase AI" (https://yourfxcase.com).

A small AI video-generation platform that I'm building and improving independently.

❤️ A Small Reality About My Startup

I want to share something that most startup posts don't talk about.

My website isn't receiving thousands of visitors. Not even close.

Some days, the traffic is genuinely disappointing.

When you're working on a product independently, managing GPU expenses, fixing technical issues, improving generation speed, and trying to reach people, it can feel overwhelming.

But something encouraging is happening.

The people who discover my website aren't always just scrolling past it.

They're creating accounts, experimenting with AI video generation, and showing genuine interest in the platform.

And seeing people actually use and enjoy something I built makes those difficult days feel worthwhile.

Maybe I don't have a massive audience yet.

But those early users give me something more valuable than impressive-looking numbers.

They give me a reason to keep building.

🌺 Why Durga Puja Means Something Special to This Project

Durga Puja reminds us of creativity.

Every pandal starts as an idea. Every sculpture begins with clay. Every beautiful decoration is the result of countless hours of someone's imagination and hard work.

Technology is not so different.

We begin with an idea, face challenges, make mistakes, and slowly bring something meaningful to life.

This Durga Puja, I wanted to connect that traditional Bengali creativity with modern AI filmmaking.

That's why I created this festive campaign:

"From Chaos to Cinematic."

Because whether you're a professional artist, a student, or someone with a beautiful photograph, your imagination deserves the chance to become a story.

🚀 Still Building. Still Learning.

YourFXCase AI is far from perfect.

There are performance challenges, GPU costs, technical limitations, and plenty of improvements still ahead.

I'm not pretending to have built the next big AI company.

I'm simply an independent creator trying to make AI video generation more accessible, one improvement at a time.

And if even a few people can use my platform to bring their ideas to life, I consider that meaningful progress.

🙏 A Small Request to the DEV Community

If you're a developer, AI enthusiast, filmmaker, or someone who enjoys experimenting with new technology, I'd love your feedback.

Try the platform: https://yourfxcase.com

Tell me what works, what doesn't, and what you think I should improve.

I'm especially interested in hearing from other independent builders who have experienced the challenge of building something with limited resources.

Because sometimes, all a small startup needs is a little support from the right community.

শুভ শারদীয়া! 🌺

From a small corner of Bengal, with creativity, technology, and a lot of hope.

— An independent creator building YourFXCase AI

Suggested DEV tags: "#showdev" "#ai" "#indiehackers" "#startup"

I'm a CGI Artist From India, Not an AI Researcher — So I Built a Free AI Video Generator

Swarnabha Mitra — Sat, 10 Oct 2026 04:44:45 +0000

I'm a CGI Artist From India, Not an AI Researcher — So I Built a Free AI Video Generator

🚀 Meet YourFXCase AI — An Ad-Supported AI Video Generator

🌐 Try it here: "https://yourfxcase.com" (https://yourfxcase.com)

A platform I'm independently building to make AI video generation more accessible, especially for creators who cannot afford expensive monthly subscriptions.

You can experiment with text-to-video, image-to-video, and short AI-generated videos, with audio capabilities depending on the generation model.

The platform offers limited free access supported by advertisements.

But here's the interesting part: I'm not originally an AI developer. I'm a CGI artist.

And this is the story behind why I decided to build it.

🎨 My Background: From CGI and Unreal Engine to AI

I'm an independent graphics designer and CGI artist from India.

My professional journey has primarily revolved around visual effects, 3D environments, animation, and digital filmmaking.

Over the years, I've developed experience working with:

Unreal Engine — Real-time rendering, cinematic environments, virtual production, and animation.
Blender — 3D modeling, materials, animation, and CGI workflows.
DaVinci Resolve — Editing, color grading, and post-production.
Adobe After Effects and Premiere Pro — Compositing, motion graphics, and video editing.

Most of my creative work involves taking an idea and transforming it into something visually believable.

And honestly, I love that process.

However, the creative industry is changing incredibly fast.

Something that once required complex 3D setups, animation, rendering, and compositing can sometimes be prototyped using generative AI within minutes.

That's when I started asking myself:

Instead of only using AI tools, why not try building a platform that makes these tools accessible to other creators?

🤖 AI Is Already Becoming Part of Traditional CGI

This is something I've noticed while working with Unreal Engine and modern rendering technologies.

AI isn't simply replacing traditional graphics pipelines. It's increasingly becoming part of them.

Take NVIDIA DLSS as an example.

Technologies like AI-powered super-resolution, frame generation, and ray reconstruction demonstrate how machine learning can improve graphics performance and image quality.

And these technologies are already available through integrations for Unreal Engine.

Of course, AI-assisted rendering and fully generative video are different technologies.

But both demonstrate something important:

The future of digital content creation will likely combine traditional CGI skills with intelligent AI-powered workflows.

I believe artists who understand lighting, composition, animation, camera movement, and storytelling will still have an important role in that future.

AI should expand our creative possibilities, not eliminate our creativity.

🇮🇳 Why I Wanted to Build Something From India

India has a growing technology ecosystem, with talented developers, researchers, startups, and creative professionals.

But I believe we still have enormous opportunities to build more accessible, creator-focused AI platforms.

Many of the popular AI video generation services can become expensive for people who are just experimenting.

For independent artists, students, and small creators, even a relatively small subscription can be difficult to justify.

I wanted to explore one simple question:

Can we make AI video generation available at little or no direct cost to the user while keeping the platform financially sustainable?

That's how the experiment started.

I didn't begin with a large team, investors, or a massive cloud infrastructure budget.

I started with my existing creative background, available AI technology, and a willingness to experiment.

And eventually, that experiment became YourFXCase AI.

🎬 What Exactly Is YourFXCase AI?

"YourFXCase AI" (https://yourfxcase.com) is a browser-based AI video generation platform focused on affordable access to generative video technology.

The idea is straightforward:

Visit the website and access the available generation tools.
Describe your creative idea or provide a reference image.
Submit a video generation request using your available credits.
Let the AI process your request.
Preview and download the resulting video.

The platform is being developed around LTX-based video generation technology.

I'm experimenting with inference optimization, GPU resource management, model configurations, and different ways to make the system more efficient.

My focus isn't simply generating videos.

It's making the entire experience affordable and practical.

💡 An Important Clarification: I Didn't Create the Foundation AI Model

I want to be completely transparent about this.

YourFXCase AI is my platform, but the underlying foundation video model is not something I trained or invented.

I'm building around existing AI technology, integrating the inference workflow into a web application, and working on the infrastructure needed to deliver video generation to users.

There's a major difference between developing a new foundational AI model and building a useful product around existing models.

Both involve different challenges.

Training a video foundation model from scratch requires extensive datasets, research expertise, infrastructure, and enormous computational resources.

As an independent creator, that isn't realistic for me right now.

But if this project becomes successful enough in the future, I would love to explore developing original AI technology with a larger team.

For now, my goal is much simpler:

Build something useful. Make it accessible. Learn from real users. Improve continuously.

💰 How Can an AI Video Generator Be Free?

This is probably the biggest challenge of the entire project.

AI video generation is computationally expensive.

GPUs cost money. Model loading takes time. Video generation consumes VRAM and processing resources.

Then there are additional expenses:

Cloud infrastructure and storage
Server maintenance
Model initialization and GPU idle time
Failed generations and retries
Website operation and development

So how do you offer free video generation without losing money on every request?

My experiment is based on an ad-supported free-credit system.

Instead of requiring every user to purchase a monthly subscription, the platform offers limited free generation access supported by advertising.

The idea is to use advertising revenue to help cover GPU and operational costs.

If the economics eventually work, the platform can become sustainable while still offering free access.

I also hope to introduce affordable paid options for users who want additional generations, faster access, or more models.

I don't believe unlimited free GPU generation is financially realistic for a small independent platform.

But limited free access with a transparent business model might be.

And that's what I'm trying to find out.

⚙️ The Engineering Challenges I'm Facing

As a CGI artist, building an AI SaaS platform has introduced me to challenges that are very different from creating visual effects.

Some of the biggest areas I'm currently working on include:

GPU efficiency: Finding the right balance between generation quality, VRAM consumption, inference speed, and operational cost.

Queue management: Handling multiple generation requests without overwhelming the available GPU.

Cold starts: Managing model loading and GPU startup delays without paying for unnecessary idle time.

Generation consistency: Improving motion quality, prompt following, and character stability.

Sustainable scaling: Making sure increased usage doesn't automatically translate into unsustainable infrastructure expenses.

I'm not claiming to have solved all these problems.

In fact, the project is still evolving.

But every experiment is teaching me something new about deploying AI models in real-world applications.

🚀 My Long-Term Vision

My dream isn't limited to operating a single AI video generator.

I'd love to see more independently developed AI tools and creative technologies emerging from India.

In the future, I hope YourFXCase AI can grow beyond its current experimental stage.

That could include more generation models, improved creative controls, faster inference, affordable plans, and eventually more original AI research and development.

I also believe there's an exciting opportunity to connect generative AI with traditional CGI workflows, including Blender, Unreal Engine, and other professional creative applications.

Imagine a future where independent creators can combine procedural 3D scenes, real-time engines, AI-assisted animation, and generative video in one affordable workflow.

That's the kind of creative ecosystem I'd love to contribute to.

❤️ A Small Request to the DEV Community

I'm still learning.

I'm not running a major AI company, and I don't have a massive GPU cluster.

I'm just an independent CGI artist from India trying to build something useful with the technology available today.

If you're a developer, designer, filmmaker, student, or someone curious about generative AI, I'd genuinely appreciate your feedback.

🎬 Try YourFXCase AI: "https://yourfxcase.com" (https://yourfxcase.com)

I'd especially love to hear your thoughts on three things:

Would you prefer limited free AI video generation supported by ads, or a low-cost subscription?
What matters most to you: speed, quality, or creative control?
What features would make an AI video generator genuinely useful for your workflow?

If you've built something similar, I'd love to learn from your experience too.

Sometimes, building something meaningful starts with a simple experiment.

And this is mine.

This is a follow-up to my previous DEV Community article: "I Built a Free AI Video Generation Website with Audio — Here's How" (https://dev.to/swarna_fx/i-built-a-free-ai-video-generation-website-with-audio-heres-how-3ac8).

Disclosure: I'm the creator of YourFXCase AI. The platform uses existing AI models, offers limited ad-supported free access, and is a work in progress. The platform itself is not being presented as open source.

Thanks for reading and supporting independent creators! ❤️

I Built a Free AI Video Generation website with Audio — Here's How

Swarnabha Mitra — Thu, 08 Oct 2026 21:30:25 +0000

I Built a Free AI Video Generation website — Here's How It Works

AI video generation is getting more powerful, but using advanced video models can still be expensive, especially when you just want to experiment with a few creative ideas.

As someone working in VFX, 3D, and Unreal Engine, I wanted to explore a simple question:

Could I build an AI video generator that people can try for free while keeping GPU costs manageable?

That's what led me to build YourFXCase AI.

It's a free-to-start, ad-supported AI video generation platform designed for creators, filmmakers, developers, and anyone interested in experimenting with generative video.

Here's what I'm building, the challenges I've encountered, and what I've learned so far.

🎬 What is YourFXCase AI?

YourFXCase AI is a browser-based video generation platform that aims to make generative video more accessible.

The experience focuses on:

Text-to-video: Turn written prompts into short AI-generated videos.
Image-to-video: Animate a reference image.
AI-generated audio: Create clips with accompanying audio.
Free generation credits: Experiment without immediately purchasing a subscription.
Ad-supported access: Help cover expensive GPU inference costs.

The goal isn't to promise unlimited free generations. It's to make experimenting with video AI easier and more affordable.

🧠 The AI Technology Behind It

I'm working with LTX-based video generation, including distilled model variants, and testing different inference settings.

One of my biggest challenges is finding a practical balance between:

Generation speed × Video quality × GPU cost

High-quality AI video generation requires significant computation.

I'm experimenting with an NVIDIA RTX A6000 with 48 GB of VRAM and evaluating how different inference configurations affect performance.

Some of the areas I'm testing include:

BF16 versus quantized inference
GPU memory usage
Model loading and warm-up time
Short-duration video generation
Resolution versus generation speed

A configuration that produces excellent video quality isn't necessarily affordable for a free service.

⚡ Challenge 1: GPU Cold Starts

Running a GPU continuously can become expensive when nobody is generating videos.

But shutting it down after every request creates another problem: model loading and environment restoration take time.

That creates a tradeoff between keeping the GPU warm for faster generation and shutting it down during inactivity to control costs.

I'm exploring how idle shutdown, model snapshots, and warm-up procedures can make that experience more practical.

🔄 Challenge 2: Handling Multiple Users

Video generation tasks consume substantial GPU memory.

If too many users request generation simultaneously, GPU memory limits can become a problem.

The architecture needs to account for:

Accepting generation requests.
Validating user credits.
Placing requests into a queue.
Processing jobs within available GPU capacity.
Returning completed videos.
Handling failures without unfairly charging users.

I'm focusing on predictable generation and reliable job handling rather than promising unlimited concurrency.

💰 Challenge 3: Making Free AI Video Generation Sustainable

GPU time isn't free.

A simple starting point for estimating inference cost is:

GPU cost per generation = GPU hourly price × generation time in seconds ÷ 3,600

But real operating costs also include idle GPU time, model warm-up, storage, retries, and infrastructure.

That's why YourFXCase AI uses a free-credit model supported by advertising.

Rather than hiding the tradeoff, I want users to understand why free generations have limits.

The long-term goal is to offer both limited free access and affordable paid options.

🎨 Why I'm Building This

My background is in visual effects and digital content creation.

I think AI-generated video can become a useful creative tool for:

Experimenting with cinematic shots
Animating still images
Visualizing story ideas
Creating short-form content
Testing creative concepts before full production

But these models still struggle with things such as character consistency, complex movements, and following detailed instructions.

Building this platform has been as much about understanding those limitations as it has been about making the technology accessible.

🚀 What's Next?

I'm continuing to work on:

Faster video generation
More reliable GPU infrastructure
Better prompt consistency
Improved user experience
Additional model options
More sustainable free usage

This is an evolving project, and there are still plenty of challenges to solve.

🙌 I'd Love Your Feedback

If you're interested in generative AI, GPU inference, or building small SaaS applications, I'd love to hear your thoughts.

What matters most to you in an AI video generator — generation speed, visual quality, audio, or affordability?

You can explore the project here:

"YourFXCase AI — AI Video Generator" (https://yourfxcase.com)

Disclosure: I'm the creator of YourFXCase AI. The platform offers limited free generation credits supported by ads. The application itself is not currently presented as an open-source project.

Thanks for reading! 🎬[]