ElevenLabs has unveiled 13 new products and upgrades at its New York Summit, in its biggest move yet from voice toolmaker to enterprise AI platform. The package, announced on November 11, includes the ElevenAI workplace app, custom models tuned on each customer’s own data, and Scribe v3, a transcription model the company calls its most accurate yet. Most of the new items are previews or early access, not finished products.
Scribe v3 leads the launch
Scribe v3 is the centrepiece. The speech-to-text model covers 102 languages and, on ElevenLabs’ own evaluation, makes 27% fewer word errors than Scribe v2. Its real-time sibling, Scribe v3 Live, labels speakers as they talk for the first time and makes 43% fewer errors than its predecessor.
Both models are in the company’s API, and prices are unchanged at $0.20 an hour for Scribe v3 and $0.22 for Scribe v3 Live. A healthcare-tuned version, Scribe v3 Medical, recognises clinical terms — the company’s example is telling the tuberculosis drug cycloserine from the immunosuppressant cyclosporine — and a financial-services model is in development.
The wider Summit package is about memory and customisation. The new Interaction Engine orchestrates voice and language models across a conversation, agents can now remember returning customers and run routines that automate follow-up work, and businesses will be able to fine-tune models on their own data. A partner programme and marketplace open the platform to outside developers and resellers.
Custom models are the most enterprise-facing item, and also the least available. Fine-tuning on a customer’s own data rolls out to selected enterprises first, beginning in the coming months, with no pricing disclosed. The company is starting with models tuned for single industries (healthcare first, financial services next) rather than opening customisation to everyone at once.
ElevenAI is the most visible new product. The app listens to conversations, including in-person meetings, and turns them into notes, summaries and action items, with dictation and meeting notes running on Scribe v3. Free early access opened with the announcement.
The partner programme is the quieter half of the strategy. ElevenLabs is opening a marketplace where developers and resellers can build on its models, and the Nvidia deal puts the Nemotron 3.5 Super VL model inside ElevenAgents as the engine behind voice agents. Both moves point the same way: ElevenLabs wants to be the voice layer other companies build on.
Our Read
Read plainly, ElevenLabs is following the same playbook as every voice company that wants enterprise money: the model is no longer the product, the platform around it is. Scribe v3’s accuracy claims are the company’s own numbers, and the first outside test is promising rather than conclusive.
Voice Arena ran Scribe v3 Live over 70 hours of calls and ranked it first in Romanian and Indian English and third in US English. That is the kind of evidence enterprises will want before they trust a transcription model with their meetings, and ElevenLabs’ audio launch followed the same pattern, with claims first and independent proof catching up later.
Why It Matters
It matters because transcription is the unglamorous backbone of the AI workplace: every meeting summary, every call-centre transcript and every voice agent starts with getting the words right. A model that is genuinely more accurate, and that knows who is speaking in real time, changes what companies can automate.
It also matters because ElevenLabs is no longer selling a tool, it is selling a platform — and the enterprise buyers it wants already buy from Microsoft, Google and OpenAI. The Summit’s real test is whether custom models and agent memory are enough to win deals against those giants, or whether ElevenLabs stays the voice supplier inside other companies’ products.
Background
ElevenLabs has shipped three new voice models in the last month: Scribe v3 for transcription, Eleven v4 for expressive text-to-speech and Eleven Realtime for fast speech-to-speech agents. The Summit also showcased enterprise customers, including Yahoo’s voice-powered AI assistant and a front-desk agent at the Westgate hotel group, as the company pushes beyond creators into the contact centre.
Source: ElevenLabs’ announcement, via The 9th Estate and CellCog’s read of the Scribe v3 launch.
0 Comments