AI Audio Models Are Dying. ElevenLabs CEO Just Admitted It

ElevenLabs built an empire on AI voice technology. Now its own CEO says those models won’t matter much longer.

Mati Staniszewski dropped this bombshell onstage at TechCrunch Disrupt 2025. His company spent years perfecting audio AI. But he believes the technology will become commoditized within two years. That’s a stunning admission from someone whose business depends on model superiority.

So why keep building them? The answer reveals how quickly the AI landscape is shifting.

The Two-Year Window

Staniszewski sees a narrow opportunity ahead. Right now, model quality still matters enormously. Bad AI voices kill user experiences. Clunky interactions frustrate customers. Poor audio ruins applications.

That gives companies like ElevenLabs a competitive edge today. Their researchers cracked difficult architecture problems. Their voices sound natural. Their technology scales reliably.

But this advantage won’t last forever. Other companies will solve these same problems soon. Plus, the performance gaps between top models keep shrinking. What seems like a breakthrough today becomes standard tomorrow.

ElevenLabs plans to maximize this window while it’s open. The company will keep pushing model development for the next year or two. After that? The real competition begins.

Multi-Modal Models Change Everything

The future isn’t just better audio. Instead, it’s about combining different AI capabilities together.

Staniszewski pointed to Google’s Veo 3 as an example. That system generates audio and video simultaneously. Other emerging models fuse audio with large language models for conversations. These multi-modal approaches create entirely new possibilities.

ElevenLabs wants to be part of this shift. The company plans partnerships with other AI firms. It will work with open source technologies. The goal is combining ElevenLabs’ audio expertise with other companies’ strengths.

This strategy makes sense if single-purpose models are dying. Why compete on audio quality alone when customers want complete solutions? Better to integrate deeply with other technologies now.

Different Models for Different Jobs

Even as models commoditize, users will still need variety. Staniszewski acknowledged this reality.

Companies building reliable, large-scale applications won’t use one model for everything. Instead, they’ll pick specialized models for specific tasks. One model might excel at natural conversation. Another handles multiple languages better. A third optimizes for speed.

So the commoditization doesn’t mean all models become identical. Rather, the differences become less dramatic. Customers can choose based on specific needs instead of accepting whoever has the best overall technology.

This creates opportunities for companies that understand real-world use cases. Pure model performance matters less than solving actual customer problems.

The Apple Strategy for AI

Staniszewski offered a revealing comparison. He said the magic of Apple came from combining hardware and software seamlessly. ElevenLabs wants to do the same thing with AI models and applications.

Building great models isn’t enough on its own. Creating useful applications matters just as much. The companies that win will excel at both.

That’s why ElevenLabs focuses on practical implementations alongside model development. Voice cloning, text-to-speech, audio generation – these applications turn raw AI capability into customer value. Without applications, even the best models sit unused.

This dual focus creates long-term defensibility. Sure, competitors will match ElevenLabs’ model quality eventually. But building great products requires different skills entirely. Understanding customers, designing interfaces, optimizing workflows – these capabilities take time to develop.

What This Means for AI Companies

Staniszewski’s comments reveal uncomfortable truths for the entire AI industry.

First, model differentiation is temporary. Companies pouring billions into training runs shouldn’t expect lasting advantages. Technical superiority fades faster than most founders want to admit.

Second, applications matter more than infrastructure. The real value comes from solving specific customer problems, not from having the most impressive benchmarks.

Third, partnerships and integration trump independence. No company can build everything customers need. The winners will collaborate effectively instead of trying to own entire stacks.

These lessons apply beyond just audio AI. Every AI company faces similar dynamics. Model quality improves relentlessly across the board. Competitive moats based purely on technology erode quickly.

The Commoditization Timeline

Two years feels aggressive for complete commoditization. But Staniszewski has direct visibility into AI research progress. He sees what’s coming in development pipelines. He knows how fast quality gaps are closing.

That timeline should worry AI companies still focused primarily on model development. If the window is truly that short, there’s limited time to establish other advantages. Companies need applications, customer relationships, and brand recognition yesterday.

Meanwhile, customers should feel optimistic. Commoditization means better technology becomes accessible to everyone. Prices drop. Quality improves. Innovation accelerates as competition intensifies.

Building for the Post-Model World

ElevenLabs is hedging its bets. The company will keep advancing models while simultaneously building applications and partnerships. That diversified approach makes sense given the uncertain timeline.

Other AI companies should consider similar strategies. Pure model plays look increasingly risky. Better to develop multiple sources of value now while model quality still provides breathing room.

The transition won’t happen overnight. Models will remain important for the next year or two. But companies betting their entire future on model superiority are making a dangerous assumption.

Staniszewski essentially admitted his company’s core advantage is temporary. That’s either brutally honest or strategically smart – probably both. Now the question becomes whether ElevenLabs can successfully transition before commoditization arrives.

The clock is ticking. Two years goes fast in AI.