VoiceHero

• 

  Voice Hero

• 

  What if your website could talk?

🎯 What Inspired This Project

Teams spend real money to attract the right people… and then send everyone to the same generic homepage.

A super‑specific visitor clicks:

“AI voice agents for ecommerce support teams”

…and lands on:

“All‑in‑one platform to boost productivity.”

That first impression doesn’t say:

• “I know who you are”
• “I know why you came”
• “I respect your time”

So we asked:

What if your homepage spoke to visitors the way a great salesperson would?

Not with chat bubbles in the corner, but with:

• A 20–30 second spoken pitch
• In the visitor’s language
• Tuned to their segment (new, returning, ad traffic, etc.)
• In a voice and tone that fits the brand

That’s the core idea behind Voice Hero:

• We scrape and understand your landing page.
• Gemini writes short, spoken pitches per segment.
• ElevenLabs turns them into natural voice‑overs.
• The widget plays the right pitch for the right visitor, instantly.

Instead of endlessly A/B testing headlines, you get a voice‑first, segment‑aware intro that evolves with your content.

🔧 How We Built It

• Firebase
  Auth, Firestore for projects/pages/scripts, hosting the dashboard and widget.

• Vertex AI / Gemini
  Takes a cleaned snapshot of each landing page + segment definition, then generates short, conversational scripts optimized for a spoken pitch.

• ElevenLabs
  Turns approved scripts into pre‑generated audio, with different voices per brand or segment.

On the dashboard, a website owner can:

1. Connect a domain and pick which pages to enable.
2. See the extracted content for each page.
3. Generate scripts per segment with Gemini, edit, and approve.
4. Generate and preview voice‑overs with ElevenLabs.
5. Drop a single <script> tag on their site to activate the widget.

Visitors then hear a tailored intro instead of reading the same static headline as everyone else.

📚 What We Learned

• Voice changes the first impression.
  A 20‑second spoken pitch feels more human than a block of text—especially when the voice is high‑quality and on‑brand.

• Segmentation beats over‑personalization.
  Per‑segment scripts (new vs returning vs ad traffic, language, page type) give most of the value without per‑user complexity.

• AI’s real work is orchestration.
  The magic isn’t just calling Gemini or ElevenLabs—it’s wiring them together in a way that:

  • keeps latency low (pre‑generate where possible),
  • keeps cost under control,
  • stays aligned with live page content as it changes.

In short: we wanted to give homepages a voice that fits who’s listening, not just who’s writing the copy.

Built With

• elevenlabs
• firebase
• firestore
• github
• next
• pnpm
• react
• shadcn
• turbo
• typescript
• vertex