The Challenge
AI-generated speech in 2022 was easily identifiable as synthetic — robotic, emotionless, and uncanny. The technology was useful for basic text-to-speech (GPS navigation, screen readers) but nowhere near good enough for creative applications like audiobooks, film dubbing, or content creation.
Existing voice AI companies (Amazon Polly, Google TTS) focused on functional quality — "good enough" for practical applications. No one was building voice AI that could convey emotion, maintain character consistency, or match the quality bar of professional voice actors.
The Approach — Tools in Action
This crystallized the product requirements:
- Emotional range (not just flat reading)
- Voice cloning (match a specific person's voice)
- Multilingual support (any language, any accent)
- Real-time generation (fast enough for interactive use)
- Discover: The worst voice AI experiences were dubbing (lost emotion), audiobooks (monotone), and accessibility tools (robotic)
- Define: The core problem was emotional fidelity — AI could produce words but not feelings
- Develop: Novel neural network architectures that modeled prosody (rhythm, stress, intonation) as a primary feature, not an afterthought
- Deliver: API-first launch, targeting developers and content creators
- Voice cloning (highest reach among creators) → launch first
- Multilingual dubbing (massive impact on global content) → second priority
- Real-time conversation (complex but transformative) → third priority
The Outcome
ElevenLabs grew explosively:
- Reached $100M+ ARR within 2 years of launch
- Valued at $3B+ — one of the fastest-growing AI startups
- Used by publishers, game studios, film companies, and millions of individual creators
- Made AI audiobook narration commercially viable — a market transformation
- Supports 32 languages with natural-sounding speech
- The technology is now used to dub Hollywood content — solving the exact problem the founders experienced as children
ElevenLabs proved that personal frustration, combined with AI capabilities, can identify massive market opportunities.
Key Takeaway
The best products often come from founders solving their own problems. Working Backwards turns personal frustration into a clear product vision — and the Double Diamond process ensures you solve the right problem at the right quality level.
Tools Used in This Story
Working Backwards
Problem SolvingStart from the ideal customer outcome and work backward to build the right thing
Double Diamond
Problem SolvingNavigate from problem to solution through divergent and convergent thinking
RICE Scoring
Decision MakingPrioritize ideas and features with a simple scoring framework