More Voices. More Markets.

Every great piece of audio starts with someone hearing it in their head before it exists — the right warmth in a voice, a pause in exactly the right place, an ad that sounds like it belongs to the market it's running in rather than being translated into it. Getting from that idea to a finished, broadcast-ready asset is where most of the real work happens. And it's rarely the model alone doing the heavy lifting.

This week, ElevenLabs shipped v4 — a step forward for the model, now covering 90+ languages with exciting new regions like Brazil and Japan, and finer control over delivery. It's one more sign of how fast this category keeps evolving.

That's exactly how we think about model partnerships: by orchestrating across every major voice, music and sound model, AudioStack can always route a brief to whichever model offers the highest quality for that job — the right voice, for the right region, for the right brief. As ElevenLabs keeps expanding into new languages and markets, that ceiling keeps rising, and our agents get another strong option to cast from.

A voice is not a production

Here's the part that's easy to miss when a new voice model ships: the model gives you a performance. It doesn't give you the brief.

Someone still has to decide which voice fits this brand, in this market, for this audience. Someone still has to write the script so it's built to be performed — not just narrated. Someone has to hit the duration to the second, lay music and sound design underneath, run QA against loudness and brand and language, and package the output to spec for wherever it's actually running.

That's the layer AudioStack operates in. Every brief that comes through our platform gets cast, written, timed, mixed, and delivered by a coordinated set of agents — the production chain that normally takes seven different specialists to run. ElevenLabs' models are one of the best inputs into that chain. v4 just made that input better: wider native-language coverage for our casting agent to draw on — including strong gains in markets like Brazil and Japan — and finer-grained delivery control for our direction agent to work with.

Scaling personalization without losing the performance

Dynamic Creative Optimization is where this compounding shows up fastest. The whole premise of DCO is variation at volume — the same core ad, adapted across markets, audiences, and placements, often running into the thousands of variants for a single campaign.

That's easy to do badly. It's much harder to do without every variant sounding like a slightly different flavor of the same generic read. In our work with Lowe's, AudioStack's Dynamic Creative pipeline produced more than 20,000 hyperlocal ad variants in three days — each one cast, scripted, and timed for its specific local market, not just swapped in with a different store name.

v4's expanding language and accent coverage makes that kind of scale sharper, not just faster. With more markets covered natively, our agents can match a hyperlocal variant to the accent and cadence of its actual market, rather than the closest approximation available. Volume and authenticity stop trading off against each other.

Production speed as a sales tool

The other place this shows up is subtler: production speed changes what a sales conversation can do.

Prisa, one of Spain's largest media groups, uses AudioStack's Instant Creative to remove the production bottleneck from radio ad sales entirely. Instead of a commercial team pitching a client and waiting days for a produced demo to come back from a studio, they generate a broadcast-ready spot live, in the room, while the deal is still being discussed.

That only works if the voice sounds finished, not like a placeholder. v4's growing coverage of languages and regional accents raises the ceiling on what “live, in the room” can sound like — a demo built for that specific audience, not a generic read that's merely close enough.

The orchestra just got a better lead voice

None of this is about whether a voice model is good enough on its own — v4 makes a strong case that it is. It's about what happens once that voice is handed to a system built to direct it: cast for the brief, written to be performed, timed to the second, mixed, and shipped to spec, at whatever volume the business needs.

v4 is now being gradually rolled out across the AudioStack platform. Book a demo to see what AudioStack's agents can do.