Staniszewski: Most remaining AI audio model problems are architectural
Mati Staniszewski · Building One of AI’s Fastest-Growing Companies | Mati Staniszewski, ElevenLabs · Sep 9, 2026 · at 15:18
Mati Staniszewski, co-founder and CEO of ElevenLabs, explains why his research team focuses specifically on audio model architecture rather than general multi-modal AI.
0:00 / 0:11exact quote · 11.5s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“On audio, we think a lot of the problems that still exist are on the architecture side. We want to be solely focused on solving those architecture problems so you can actually get the best quality out there.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Mati Staniszewski
Staniszewski: Deploying forward deployed engineers without product feedback loops equals services
“And I feel like the second part is always frequently missed was like first, okay, if these are, you know, doing effectively a lot of the hard integration work to solve the problem, but if you don't do the second step of how you actually learn the expertise and…”
Staniszewski predicts conversational AI will deliver human-like dialogue within 12 months
“So, so I think 12 months from now, what will happen is, is similar as we are speaking. It's, I think the technology will be able to understand us at an incredibly better way, both our knowledge, both the emotional part of that conversation and deliver, deliver…”
Staniszewski claims ElevenLabs built the first human-quality speech model in 2022
“Built all the audio models, starting with model to produce speech text to speech model. And that was that 20, 20 to the first model that could finally cross that human like quality.”
Staniszewski: ElevenLabs will not build text-to-video foundation models from scratch
“Still a conscious decision we are taking is not creating the model from scratch. So like true text to video or like, you know, VO free type models that exist. We want to be at that intersection where we know the audio can give you the unique advantage”
Staniszewski: ElevenLabs embeds engineers in non-technical teams like legal and ops
“In all the small teams that we have, in all the teams that we have, we bring, even non-technical teams, we bring engineering talent in those teams. So our talent team, our ops team, our legal team, all of that will have engineers that help both. Automate some …”
Staniszewski: Voice agents extract far more inbound lead data than web forms
“One, people are both making their things easier. They enjoy it more, but two, they leave a lot more information than they ever would for the dropdown form. They tell us so many about the problems, about the wider set of use cases, so we can connect them better…”