Solvd

What it does

Solvd automatically transforms any physics problem into a concise, step‑by‑step animated walkthrough, solving the frustration students face when long text and static diagrams make it hard to visualize and understand complex concepts.

Your inspiration

We stumbled on this problem while building tutoring agents with Google’s ADK: our bots could solve physics questions but only offered long text and static images, leaving students lost. Meanwhile, experimenting with Manim showed us how powerful real‑time, narrated animations could be. We realized no one had married an agentic interface with on‑demand visuals, so we prototyped a system that turns any student query into a short, voice‑over animation—and Solvd was born.

How it works

Solvd works in four straightforward stages: 1. Read & Break Down You enter a physics problem and Solvd “reads” it, splitting it into clear steps (e.g., draw the free‑body diagram, apply Newton’s law, solve for acceleration). 2. Storyboard Creation For each step, it builds a simple script—“show the block on a slope,” “draw force arrows,” “display the equation”—which an animation tool (Manim) uses to sketch shapes, arrows and numbers. 3. Cloud Rendering These mini‑storyboards are sent to cloud servers that render them into short video clips in parallel, keeping wait times low. 4. Voice‑Over Sync A text‑to‑speech engine generates narration (“First, here’s the block on the incline…”) synced to the animation. Within seconds, Solvd delivers a concise, narrated animation in your browser—no more flipping through pages of text.

Design process

In just six weeks, we transformed a glimmer of inspiration into a fully functional beta: in Week 1, we sketched a Manim‐based proof of concept that automatically animated a simple free‐body diagram from a textbook problem; during Weeks 2–3, we recruited our first cohort of 50 engineering students, collected detailed feedback on usability and painfully long render times (averaging 2–2.5 minutes), and rolled out interactive controls—pause, rewind, step highlights—and a clear progress bar; in Weeks 4–5, we tapped into the expertise of leading ed‑tech and animation mentors who helped us refine pacing, visual hierarchy, and narration scripts, while our engineering team introduced request queuing, partial result caching, and lightweight front‑end optimizations that shaved off 30–45 seconds per video; and by Week 6 we began training and fine‑tuning our proprietary AI model to produce Manim scene descriptions in mere seconds, positioning us to reduce end‑to‑end render times to under 15 seconds as we gear up for wider beta rollout and continued platform enhancements.

How it is different

What sets Solvd apart is its seamless fusion of conversational AI and real‑time, narrated animations: unlike other platforms that offer only static diagrams or text walk‑throughs, Solvd’s agent reads your exact question, breaks it into steps, and on demand spins up a custom Manim animation with synchronized voice‑over—all in under 15 seconds once our proprietary model is live. We’ve built a lightweight AI to generate the scene descriptions directly, use cloud‑based parallel rendering with smart caching, and layer in interactive controls (pause, rewind, step jump) so learners can explore each concept visually and conversationally—something no other physics tutor has combined at this speed and depth.

Future plans

In the next phase, we’ll apply cognitive load theory by minimizing extraneous load and optimizing intrinsic complexity. Each animation will be broken into micro‑steps with clear scaffolding and brief interactive quizzes between segments. Real‑time feedback on mental effort will let our AI adjust narration pace and visual detail per learner. Over the coming quarter, we’ll refine these adaptive features, introduce spaced‑repetition boosters, and expand coverage across the full physics curriculum.

Awards