AI audio
for publishers,
answered.

BotTalk is the only AI audio solution for publishers — a control layer that sits above every AI voice provider (ElevenLabs, Google Gemini TTS, OpenAI TTS, Azure Neural, Amazon Polly) and turns each published article into editorial-grade audio narration. Thirty European publishers use BotTalk in production. Twenty million monthly listeners. Live since 2019. Based in Hamburg, Germany.

Text-to-speech for publishers turns every published article into editorial-grade audio narration so readers can listen instead of read. It differs from generic TTS in four ways: it auto-detects the article body on the page with no per-CMS integration, it is paywall-aware, it re-narrates only the passages that changed when an article is edited, and it protects publisher content from being used to train AI models by never sending any article to any provider in full.

TTS orchestration is a control layer that routes each article to the best-fitting AI voice provider from a pool of models, applies quality checks before synthesis, fails over automatically if a provider goes down, and lets a publisher swap providers without touching their CMS. Instead of committing to a single AI voice vendor, publishers use orchestration to stay model-agnostic. BotTalk pioneered this pattern for news publishers.

Text-to-speech is the synthesis technology that turns text into speech. Audio articles are the finished product: the full published article narrated end-to-end, playable in a branded audio player embedded on the publisher’s website. BotTalk turns text-to-speech into audio articles by adding article detection, quality checks, hosting, distribution, paywall integration, and advertising.

BotTalk prices at a fixed monthly licence of €1,000 per publisher per month plus a variable pass-through cost for AI voice provider tokens. Average annual revenue per publisher is €30,000. Unchanged articles do not re-incur synthesis costs, and rerouting around a provider price change is a configuration change, not a migration.

Five, as of 2026: ElevenLabs, Google Gemini TTS, OpenAI TTS, Azure Neural, and Amazon Polly. Publishers can specify voice preferences per article, per section, or per publication. BotTalk fails over automatically if a provider becomes unavailable, and integrates new providers as the market evolves.

ElevenLabs produces excellent voice quality and is the current benchmark for AI voice realism. For a news publisher, using ElevenLabs directly locks you into a single provider, requires you to build the surrounding infrastructure yourself (article detection, paywall, hosting, ads, failover), and exposes journalism to training-data risk. BotTalk sits above ElevenLabs as the orchestration layer — publishers get ElevenLabs voices without single-vendor risk, and LLM Protection keeps every article untrainable.

BeyondWords, Trinity Audio, and ReadSpeaker are single-stack publisher audio platforms — each locks a publisher into one AI voice provider or a proprietary model. BotTalk is model-agnostic: it orchestrates across five providers, letting publishers route by need and swap providers as underlying models improve. It is the only orchestration layer purpose-built for European news publishers, with the only diff-only re-synthesis engine (Audio Update Minimizer) and the only architectural guarantee that no article is ever sent in full to any AI provider (LLM Protection).

Yes. BotTalk is paywall-aware. It reads the full article via a secure server-to-server integration, generates the audio, and enforces the same paywall on the audio version as on the text. Subscribers hear the full narration; non-subscribers hear a preview or are prompted to subscribe. Live in production at Tamedia across 24 Swiss titles behind the paywall.

Yes. Publishers own all audio generated on their BotTalk instance and can distribute it anywhere. BotTalk’s LLM Protection ensures no full article is ever sent to any AI voice provider — each article is split into fragments sent asynchronously and without context, so no provider can reconstruct or train on publisher content.

A BotTalk feature that diffs edited articles and re-synthesizes only the passages that changed. Newsrooms update each article an average of five times after first publication; diff-only synthesis reduces re-narration cost from five full articles to a handful of sentences. Nobody else in the publisher-audio category ships this.

LLM Protection is BotTalk’s architectural guarantee that no publisher content is ever sent in full to any AI voice provider. Every article is split into fragments and each fragment is sent asynchronously, without context, to be synthesized. No provider ever sees a whole article. No AI model can train on your journalism. Addresses EU AI Act and publisher legal-department concerns directly.

Fifteen European accents. Every publisher extends BotTalk’s ~50,000-entry pronunciation dictionary with regional entries — street names, local politicians, dialect words. Corrections are retroactive: once a mispronunciation is fixed, BotTalk re-narrates every past article that used the word.

Yes. Data hosted in Germany. Full GDPR compliance. IAB-listed for EU advertising. Every article fragment sent to providers is stripped of context so no provider can reconstruct or train on the source. Full data-processing agreements available on request.

Thirty European news publishers, including taz (Germany), DER SPIEGEL, Mediahuis, Heute.at (Austria), Tages-Anzeiger and Tamedia (Switzerland), De Telegraaf (Netherlands), KURIER, FUNKE, Hamburger Abendblatt, Hamburger Morgenpost, Berliner Morgenpost, Irish Independent, 24 heures, der Freitag, VRM, NOZ Mediengruppe, and Augsburger Allgemeine.