How I Built a Free AI Speaking Coach in Minutes Using Gemini 3
What if you could get real-time feedback on your public speaking skills — your body language, vocal variety, pacing, filler words, and more — without paying a dime or even creating an account? That's exactly the tool I built using Google's Gemini 3, and the entire experience revealed just how powerful (and accessible) AI-driven app creation has become. Here's the full story of how it came together and how you can try it yourself.
Why I Wanted to Build an AI Speech Coach
There's been a lot of buzz around Gemini 3 and its ability to let everyday users "vibe code" — essentially describe what you want in plain language and watch the AI build it for you. I wanted to put that promise to the test with something practical: a tool that could help people improve their public speaking.
Public speaking is one of those skills that benefits enormously from feedback, but most people practice in isolation. They rehearse in front of a mirror or record themselves and cringe through the playback. I wanted to create something that could act as an always-available coach — one that watches, listens, and offers structured feedback the moment you finish speaking.
What the Tool Actually Does
The AI speaking coach analyzes your speech across several key dimensions:
- Body language and gesture control — Are you using your hands effectively? Is your physical presence engaging?
- Vocal variety and tonal range — Are you monotone, or do you use pitch and inflection to keep your audience interested?
- Speaking pace — Are you rushing through your points or giving your audience time to absorb what you're saying?
- Pause usage — Strategic pauses are a hallmark of great speakers. The tool tracks whether you're using them.
- Energy levels — Does your delivery feel alive and dynamic?
- Filler word detection — The tool transcribes your entire speech and flags every "um," "uh," "like," and "you know."
For filler word detection to work, Gemini 3 had to recognize on its own that it needed to transcribe the audio in real time — a small but impressive detail that speaks to the intelligence baked into the platform.
How Fast It All Came Together
Here's the part that still amazes me: the base product was functional within 30 seconds. I described what I wanted in plain language, and Gemini 3 generated a working prototype almost immediately. From there, I spent roughly an hour fine-tuning — adjusting scoring categories, refining the interface, and ironing out bugs.
And when bugs did appear? I simply described the issue in normal, conversational language, and the AI corrected it. No coding knowledge required. No debugging sessions. Just a back-and-forth conversation that steadily improved the tool.
I want to be transparent: I am not an overly technical person. The fact that someone without a development background can build a functional, useful application this quickly is a genuine sign of where AI technology is headed.
How to Use the Tool
Using the AI speech coach is straightforward:
- Access the tool via the link provided in the original video's description. It's a web-based app — no downloads needed.
- Grant permissions — The tool needs access to your camera and microphone. Google will also warn you that it was created by a third party, which is a healthy reminder to always verify you're accessing tools from trusted sources.
- Hit "Start Speech" — Once loaded, you'll see a standby screen with a simple button. Click it, deliver your speech, and stop when you're done.
- Review your scores — The tool will generate ratings across all the categories mentioned above, along with a full transcription of your speech with filler words highlighted.
There's no sign-in required and no cost involved. You simply open the URL and start speaking.
Putting It to the Test
To demonstrate the tool in action, I delivered an impromptu story — completely off the top of my head — about an embarrassing encounter at my favorite coffee shop. After I finished, the tool scored me across every category. It flagged that my speaking pace was a bit rushed, that I could have incorporated more strategic pauses, and that my body energy had room for improvement. On the positive side, it rated my tonal range and gesture control favorably.
Was the feedback perfect? Not entirely — this is still a work in progress. But it was useful, and that's what matters. The kind of structured, immediate feedback the tool provides is exactly what speakers need to identify patterns and make incremental improvements over time.
What's Next — and How You Can Help
This is very much a living project. There will be bugs, and each time I push an update, the URL may change. But the goal is to keep iterating based on real user feedback. If you try the tool, I'd love to hear from you:
- Did it work well on your device?
- Were the scores and suggestions helpful?
- What features or improvements would make it more useful?
And if you've built something of your own with Gemini 3, I'm genuinely curious to hear about your experience. The possibilities with this technology feel wide open.
Conclusion
Building an AI-powered speaking coach in under an hour — with no coding expertise — would have sounded like science fiction just a couple of years ago. Today, tools like Gemini 3 make it a reality. While the tool isn't perfect, it represents something genuinely exciting: accessible, free, AI-driven feedback that can help anyone become a more confident and polished public speaker. Give it a try, share your thoughts, and keep practicing. The best speakers aren't born — they're built through repetition, feedback, and a willingness to improve.