feed of video options
Image of Electrical Engineering Section

I-nteractive — Prompt-to-Video that Talks Back

Inspiration

Short-form video usually feels one-way. We wanted a loop that talks back—you type a prompt, get a tailored video, respond again, and the system adapts. Think “chat,” but the replies are cinematic clips instead of text.

What it does

Prompt → Video: Enter a natural-language prompt and receive a short, thematically consistent video.
Customizable Generation: Choose settings, characters, and themes before rendering.
Guided Scripts: Optionally include an auto-generated script so users can follow along or re-shoot.
Conversational Loop: Each new prompt can build on the last video, creating an iterative story or tutorial.

How we built it

Frontend: Next.js (App Router), React, Tailwind CSS
Backend/ML: Python, PyTorch
Services: Luma AI (video generation), Fetch.ai (orchestration/agents), ElevenLabs (voice), OpenAI (planning/prompting) DataBase TinyDB

High-level flow:

User submits prompt + options from the Next.js UI.
Backend orchestrates generation (FetchAI/OpenAI for planning → Luma for video → ElevenLabs for VO), with status polling.
Electrical Section → FetechAI for determining what type of circuit is needed → Google Gemini for making the netlist and .asc file → PYLTspice for simulating and creating a circuit diagram.
Pytorch VAE model for Query search.
The resulting asset is returned to the UI for playback and iteration.

Challenges we ran into

Latency & Throughput: Generation queues caused noticeable delays.
Duration vs. Cost vs. Quality: Faster, higher-quality, longer clips spike costs—hard to optimize during a hackathon.
Tooling Glue: Wiring multiple services reliably under time pressure (auth, retries, and fallbacks).

Accomplishments we’re proud of

Shipped an end-to-end, multi-service pipeline in hackathon time.
Built a usable UI that hides complexity and lets users focus on creative iteration.
Established a scalable orchestration pattern we can harden after the event.

What we learned

Scope ruthlessly: Clear MVP boundaries kept us shipping.
Design for failure: Retries, progress states, and timeouts are table stakes for generative pipelines.
Team comms matter: Fast decisions and clear ownership beat polishing non-essentials.

What’s next for I-nteractive

Smoother UX: Better progress feedback, cancel/modify mid-render, and more intelligent defaults.
Efficiency: Model/pricing tiers, caching, and partial re-use to cut both latency and cost.
Quality: More consistent characters and styles across the conversational loop.
Safety & Rights: Clear guardrails on inputs/outputs and attribution for generated assets.

Quick FAQ (optional in submission)

Why now? Short-form creation is exploding, but iteration still feels clunky. We make it truly conversational.
Who’s it for? Creators, educators, support teams—anyone who wants rapid video iteration without editing timelines.