Inspiration
Founders practice in mirrors. Mirrors don't whisper "this TAM is fake" behind their backs—real VCs do, so we built the mirror that fights back.
What it does
Upload your pitch deck. Three AI VCs call you, ask brutal questions, and whisper their real doubts mid-conversation so you hear what they actually think.
How we built it
Next.js Pages Router feeding PDFs to Gemini for analysis, ElevenLabs TTS generating three distinct voices with pre-cached audio queues to avoid demo latency. We faked the "live" video with Framer Motion avatar pulses and overlapping whisper tracks because real-time WebRTC would have exploded.
Challenges we ran into
Synchronizing three audio personalities without crosstalk—turns out one Gemini brain controlling three voices is easier than three agents arguing. Filler word detection required exact transcript timestamps; we learned "um" is surprisingly hard to regex when people mumble at 2am.
Accomplishments that we're proud of
The whisper mechanic actually feels creepy—users physically lean in when the VCs start talking behind their backs. We shipped a zero-latency demo by pre-generating audio, proving you can fake "real-time" if you cache intelligently.
What we learned
Audio UX is 90% timing—if the pause between question and answer is >800ms, the magic dies.
What's next for One Hundred Million Dollar Startup Idea
Kill the gimmick: replace the fake VCs with actual retired angels' voice clones and real term sheet data. Then sell it to YC as their "winter batch hazing tool."
Built With
- elevenlabs
- gemini
- next.js
- typescript
Log in or sign up for Devpost to join the conversation.