BotTalk Audio infrastructure for publishers

Text-to-speech
built for
publishers.

Text-to-speech for publishers isn’t a voice tool. It’s infrastructure. BotTalk is one control layer above every AI voice engine — it turns every published article into editorial-grade narration, routes across five engines, and never goes dark. Running today across 30 European newsrooms.

Turning articles into audio is easy. Running audio as infrastructure for a newsroom is not. That’s the difference between a voice tool and text-to-speech built for publishers.

A publisher needs more than a voice. You need every article narrated automatically, a player on the page, a podcast feed off it, paywall and consent handled, ad inventory that pays for it, and an audio pipeline that keeps working when a voice engine reprices or goes down. BotTalk is the control layer that does all of that — one integration, above five AI voice engines.

One integration. Five engines. Every article.

You add one script tag. The layer auto-detects new articles from your CMS or RSS, runs them through a pre-synthesis quality engine, routes each to the voice engine that fits, and returns editorial-grade audio — on the page and in an auto-generated podcast feed. If an engine fails or reprices, the layer reroutes. Your integration never changes.

How BotTalk text-to-speech works Your article enters the BotTalk control layer, which runs a five-check quality engine and routes across five AI voice engines — ElevenLabs, Gemini, OpenAI, Azure Neural, and Amazon Polly — with automatic failover, then returns editorial-grade audio to an on-page player and a podcast feed. YOUR ARTICLE BOTTALK LAYER QUALITY ENGINE · 5 CHECKS ELEVENLABS GEMINI OPENAI AZURE POLLY ROUTE · FAILOVER · NEVER GO DARK OUTPUT → ON-PAGE PLAYER + AUTO PODCAST FEED
One script tag in. A five-check quality engine, five voice engines with failover, and a player + podcast feed out.

The five things a newsroom must control

Once audio is infrastructure, a single AI voice engine is a single point of failure. The control layer is what turns five separate engines into one dependable service — and hands the newsroom back control of the five things that matter.

  • Quality. A pre-synthesis quality engine normalizes numbers, names, and dialect before any model speaks — 50,000 dictionary entries, 15 European accents.
  • Cost. A fixed monthly licence plus pass-through tokens. When an engine reprices, you route around it instead of absorbing it.
  • Uptime. Automatic failover across five engines. Zero customer-facing outages through three documented provider incidents in the last twelve months.
  • Language. Each language routed to the engine that renders it best, so coverage expands without new integration work.
  • Brand voice. Cloned editor voices under documented, revocable consent — the voice your audience recognizes stays yours.

Comparing vendors first? Read our honest, side-by-side comparison of the best text-to-speech for news sites, and why we argue you shouldn’t bet audio on one AI voice provider.

Running today, in production

Not a pilot. The BotTalk control layer runs audio for 30 European newsrooms — including taz, DER SPIEGEL, Tamedia, heute.at, and Mediahuis — and streams to 20 million monthly listeners.

5
AI voice engines, one integration
30
European publishers live today
20M
Monthly listeners across the network
15
European accents on one policy
Pascal Vanz, Product Manager Web/App at Tamedia

“The rapid deployment and impressive engagement metrics during the POC were beyond our expectations. Expanding BotTalk to our other newspapers was an easy decision, and the premium content feature has added significant value for our subscribers.”

Pascal Vanz Product Manager · Web/App · Tamedia
Read the Tamedia case study
Felix Herkenrath, COO at Hamburger Morgenpost

“The easy integration of the audio advertising platform was a turning point for us. It seamlessly integrated into our existing systems and addressed our biggest problems with declining print ad revenues.”

Felix Herkenrath Chief Operating Officer · Hamburger Morgenpost
Read the Mopo case study

Just a play button. And everything behind it.

The reader sees one play button. Behind it, the publisher-native stack that a raw voice engine doesn’t give you:

  • CMS & RSS auto-detection — new articles narrated automatically, no per-article work.
  • Paywall & GDPR consent — handled in the layer; German-hosted.
  • IAB-listed audio ads — audio that pays for itself.
  • Auto podcast feeds — the away-from-tab surface, generated per publisher.
  • Five-provider failover — one engine’s outage is a reroute, not your silence.

Route by need. Swap any time. Never go dark.

Frequently asked

Text-to-speech for publishers, answered.

What is text-to-speech for publishers?

Text-to-speech for publishers turns every published article into editorial-grade audio narration, so readers can listen instead of read. For a newsroom it isn’t a single voice tool — it’s infrastructure: a layer that narrates every article automatically, handles the player, paywall, ads, and consent, and keeps working when any one AI voice engine changes or goes down. BotTalk provides that layer across five engines.

How do publishers add audio to their articles?

With BotTalk it’s one integration — a single script tag or API. The layer auto-detects new articles from the CMS or RSS, narrates them, and drops an audio player onto the page. No per-article work: publish an article and the audio appears. An auto-generated podcast feed ships alongside the on-page player.

Is text-to-speech for publishers GDPR compliant?

Yes. BotTalk is German-hosted, handles GDPR consent, and its audio advertising inventory is IAB-listed in the EU. Voice cloning of a named editor is used only with documented, revocable consent. The EU AI Act’s synthetic-audio transparency obligations are enforced in the layer rather than re-implemented per engine.

How long does it take to integrate?

Plug-and-play from day one — one integration, no per-engine work. Publishers in the BotTalk network go live in days, not quarters, because the routing, failover, provider contracts, and language coverage live inside the layer, not the publisher’s codebase.

Which languages and accents does it support?

BotTalk covers 15 European accents on one integration and routes each language to the engine that renders it best. Because Europe runs on 24 official languages and no single engine handles all of them well, routing across five engines is how coverage expands without new development work.

How is text-to-speech for publishers priced?

BotTalk prices by a fixed monthly publisher licence plus pass-through provider tokens — predictable and independent of any single engine’s repricing. Raw engines bill per character; single-vendor tools expose you to that vendor’s price changes. A control layer lets you route around a price move instead of absorbing it.

Hear it on your own articles

Talk to Andrey.

No deck. No pitch. Your CMS, your audience, your editorial voice. Your own articles narrated live, routing and failover walked through on the call. Thirty minutes. One call.