The spark behind this project
A few days ago, OpenAI released Sora 2, an incredibly powerful text-to-video model that can generate minutes of realistic footage from just a few lines of text.
The results were mind-blowing — and I couldn’t wait to try it out myself.
So, I decided to integrate Sora 2 into my own project, Textideo, a website I’ve been building in my spare time.
It’s now live, and users can directly create stunning short videos using the same Sora 2 model — no coding or subscription required.
Why I built this
Last year, I was laid off from my job. It wasn’t easy, but it forced me to rethink everything.
I realized that in the AI era, those who keep learning and mastering new tools will always find new opportunities.
AI tools have dramatically boosted my productivity — automating repetitive work, speeding up content creation, and helping me experiment faster.
Over time, I became confident enough to build my own AI-powered platform.
That’s how Textideo was born.
What Textideo Sora 2 does
Textideo Sora 2 lets anyone turn text and a single photo into a short AI-generated video.
You simply:
- Upload a photo
- Write a short prompt or script
- Let Sora 2 generate the video for you
- Control who can access or use your visuals
- Instantly share or embed your creation
It’s simple, fast, and accessible to creators who want to bring their imagination to life — without needing GPUs or complex setups.
The challenge behind the magic
Generative AI is powerful — but also resource-heavy.
Every “magical” video comes with real energy, water, and compute costs behind the scenes.
Training and running models like Sora 2 demand massive GPU clusters, and each second of HD video can consume as much power as running a microwave for over an hour.
That’s why I’ve focused on efficiency, modularity, and transparency in Textideo’s integration — balancing creativity with sustainability.
What I’ve learned while building it
Here are some lessons I’ve picked up during development:
- Modular architecture matters — break down features like prompt parsing, face alignment, and video encoding into reusable parts.
- Optimize gradually — build something that works first, then refine it. Batch processing, caching, and quantization help a lot.
- User control builds trust — privacy settings for user likenesses are crucial.
- Data-driven improvements — monitor latency, GPU time, and user engagement to make smarter updates.
- Start small, scale smart — don’t jump straight to clusters; test and iterate.
The results so far
- My side income has reached around $1,000 per month
- Users are excited about seeing themselves in “AI magic videos”
- I’ve grown tremendously as a full-stack and AI developer
Try it yourself
I’d love for you to give it a try and see what you can create.
👉 Experience Textideo Sora 2 here
If you’re a developer or creator, I’d really appreciate your feedback — on performance, features, or even business strategy.
Let’s see what you’ll build next. 🚀
Top comments (0)