No comments yet.
We’ve been passionate about public speaking for as long as we can remember. Personally, I (one of the creators) have loved presenting ever since 6th grade, and in college, I was president of my Toastmasters chapter. Another creator was among the highest-rated public speakers at one of the world's largest companies. Over the years, our team has seen how great speaking skills can open doors — whether it’s giving a talk, pitching an idea, or just leading a meeting with confidence.
That’s why we built SpeakBrightly.com, an AI-powered public speaking coach that helps you become a confident and effective public speaker. It does so by providing structured feedback on your delivery, clarity, and engagement. Our goal is to make high-quality speaking coaching accessible to anyone.
Here are some interesting technical tidbits for the other builders out there :) - Align feedback with explicit examples: Early on, we noticed that AI-generated feedback was too vague (even when scoring against a detailed rubric). But by forcing the evaluation to reference explicit examples from the speech, we significantly improved usefulness and accuracy. - Preprocess audio to get the best results from GPT-4o: Downsampling MP3s to 16kHz before sending them to the audio model dramatically improved transcription quality and feedback accuracy. - Process audio in two stages: We found that generating an initial report with the audio model and then feeding the transcription plus the audio model’s report into a text-only model provided the most precise and actionable insights.
We’re excited to finally share this with the world! If you’ve ever wanted better feedback on your speaking skills, we’d love for you to try it out and let us know what you think.
