UhmDetector.ai

Landing page

Inspiration

The idea for UhmDetector.ai was born from a personal struggle — my own insecurities around public speaking. I often found myself using filler words like “uhm” or “like” without realizing it, and it affected my confidence during presentations, interviews, and everyday conversations.

To make it more challenging, I also struggled with typing fluency, which made it hard to express myself clearly and efficiently when preparing speeches or responding to questions. Rather than ignore the problem, I decided to build a solution that could actively guide and train me — not only to become more mindful of my spoken words but also to improve my speech and typing clarity at the same time.

UhmDetector.ai was created as a fun, judgment-free tool to help myself (and others like me) become better communicators — one word at a time.

What it does

UhmDetector.ai is a comprehensive web application designed to help users improve their communication skills through real-time speech analysis and typing proficiency training — all in a fun, gamified experience.

🎤 Speech Analysis & Improvement Live Filler Word Detection: Instantly identifies and counts filler words like “uhm,” “like,” “so,” and “you know” while you speak.

Clarity Scoring: Provides a percentage-based score that reflects speech fluidity and confidence.

AI-Powered Suggestions: Offers intelligent feedback with alternative phrasing and speaking tips.

Practice Modes tailored for:

Elevator pitches (30 sec)

Storytelling (2 min)

Interview prep (90 sec)

Impromptu speeches (60 sec)

Speech Analytics: Tracks pace, tone, sentiment, and confidence.

Session Transcripts: Saves each speech session as a transcript, with filler words clearly highlighted for review.

⌨️ Typing Speed & Accuracy Training Practice with words, sentences, code snippets, and more.

Choose from beginner to advanced levels.

Live stats on WPM, accuracy, and mistake patterns.

Targeted exercises based on user error patterns.

Full progress tracking and improvement history.

📊 Analytics & Gamification Visual dashboards showing improvement over time.

Detailed breakdowns for both speech and typing.

Badges, milestones, and a global leaderboard to keep users motivated.

🔐 Privacy & Accessibility Local speech recognition ensures no audio ever leaves your device.

No audio is stored — only text transcripts are saved securely.

Secure email login with OTP-based password reset (via EmailJS).

Fully responsive across devices and GDPR compliant.

👥 Built for Everyone Professionals: Public speakers, educators, sales teams.

Students: Presentations, interviews, thesis defenses.

Content Creators: Podcasters, streamers, YouTubers.

Anyone: Who wants to speak and type more clearly and confidently

How we built it

We built UhmDetector.ai entirely using Bolt.new — an AI-powered platform for rapidly developing full-stack SaaS applications.

Instead of manually writing every component, I prompted Bolt's AI to generate functional UI elements, responsive layouts, authentication flows, and logic modules. This allowed me to focus on product thinking, feature design, and user experience instead of boilerplate code.

Key tools and technologies used:

⚙️ Bolt.new AI Builder — Used to scaffold pages, auth, forms, and data storage

🎤 Web Speech API — For real-time speech-to-text processing in the browser

🔢 Custom logic — For filler word detection, scoring, and session tracking

📊 State Management — For real-time updates during live speech and typing

🧠 AI Prompts — Used to build typing modes, performance dashboards, and gamification logic

💅 Tailwind CSS — For styling and consistent design language

📨 EmailJS — For secure OTP-based password resets

🔐 Bolt’s built-in auth system — For account management and data isolation

The entire app is browser-based, responsive across devices, and built in record time thanks to Bolt.new’s AI-assisted development environment.

Challenges we ran into

Real-time transcription timing: Ensuring minimal delay while detecting filler words live was tricky — especially on slower connections.

Speech accuracy: The speech-to-text engine sometimes struggled with accents or background noise, so we had to fine-tune our detection logic.

UI sync with audio: Coordinating real-time updates in the UI while keeping performance smooth took a lot of tweaking.

Keeping it fun: We wanted the experience to feel encouraging, not punishing — balancing that tone in the design and features was key.

Accomplishments that we're proud of

Built a fully functional SaaS app in a short time — from landing page to live mic analysis and leaderboard.

Made speech self-awareness accessible and enjoyable for anyone.

Created a UI that people genuinely like playing with — even outside of “serious” speaking practice.

Delivered a usable MVP that could be plugged into Zoom, classrooms, or speaking clubs in future.

What we learned

Real-time browser-based speech analysis is powerful — but needs thoughtful UX to be effective.

Gamification can make even awkward self-improvement goals (like removing “uhms”) surprisingly fun.

Bolt.new enabled rapid UI/UX iteration without diving deep into raw frontend code, which gave us more time to focus on the user experience and logic.

What's next for UhmDetector.ai

Now that the core features of speech and typing analysis are live, the next phase of UhmDetector.ai will focus on deepening the learning experience and expanding accessibility to a global audience.

🔊 Pronunciation Training We plan to introduce AI-powered pronunciation guidance, allowing users to:

Practice difficult words and phrases

Get real-time feedback on articulation and enunciation

Compare their speech to native examples and receive tailored exercises

This will make the app even more helpful for non-native English speakers, public speakers, and anyone looking to sound clearer and more confident.

🌐 Multi-language Support To reach a broader audience, we’re expanding filler word detection, typing challenges, and analytics to support:

African languages like Zulu, Xhosa, and Afrikaans

Widely spoken languages such as Spanish, French, and Hindi

This will empower users worldwide to improve communication skills in their native or second languages — making the app more inclusive and impactful.

📲 Mobile Optimization & Integrations Improve mobile experience for on-the-go practice

Explore integrations with Zoom, Google Meet, or Chrome to offer live analysis during real meetings and recordings

🧠 Smarter Feedback Introduce AI coaching insights that explain why a filler word was used and suggest behavioral or rhetorical techniques to reduce them

Add speech topic prompts and timers for realistic public speaking simulation