innFactory AI Consulting from Rosenheim, Germany advises DACH-region enterprises on GDPR- and AI-Act-compliant deployment of voice and music AI. In spring 2026 ElevenLabs expanded its portfolio in quick succession: Eleven v3 went GA in March, followed in May by Eleven v3 Conversational for voice agents and Music v2 for commercially cleared AI music.
What is Eleven v3?
Eleven v3 is ElevenLabs’ current flagship model. It generates natural-sounding speech with high emotional range and contextual understanding in more than 70 languages.
Key innovations
Audio tags
For the first time, emotional and acoustic cues can be steered directly in the text – via tags in square brackets:
[excited] We are thrilled, [whispers] that you're here today.
[sighs] After a long day…
[clapping] Well done!The model interprets tags such as [excited], [whispers], [sighs], [gunshot], [clapping] or [explosion] and adapts tone and audio accordingly.
Text-to-Dialogue API
The Text-to-Dialogue API lets you generate multi-speaker scenarios (podcasts, audio drama, training videos) in a single call – with natural dialogue dynamics between multiple voices.
Broadest language coverage
70+ languages with high quality – covering all major European languages and many smaller languages that competitors miss.
New in May 2026
Eleven v3 Conversational
With Eleven v3 Conversational, ElevenLabs closes the gap between studio quality (Eleven v3) and real-time latency (Flash v2.5). The variant inherits audio tags and multi-speaker logic from v3 but is optimised for live dialogue with voice agents. An Expressive Mode combines the model with Scribe v2 Realtime, allowing agents to infer emotion from the user’s voice and react accordingly. Turn-taking including interruption handling runs natively in the model.
Music v2
Music v2 generates complete songs including vocals and arrangement in studio quality. New features include section-level inpainting (selectively regenerate verse, bridge or chorus) and mid-track genre switching. Critical for enterprise and marketing use cases: Music v2 was trained exclusively on licensed data and is cleared for commercial use – a clear advantage over competitors such as Suno or Udio, both facing lawsuits from major labels. Pricing for Music v1 and v2 was cut by up to 50% at the same time.
Model selection by use case
| Use case | Recommended model | Why |
|---|---|---|
| Marketing videos / ads | Eleven v3 | Highest quality, audio tags |
| Audiobooks / narration | Multilingual v2 | Stable for long-form |
| Voice agents / telephony | Eleven v3 Conversational | Latency + audio tags + Expressive Mode |
| Ultra-low-latency bots | Flash v2.5 | Lowest latency |
| Multi-speaker podcasts | Eleven v3 (Text-to-Dialogue) | Multi-voice dialogue |
| Ad music / jingles / sound branding | Music v2 | Commercially cleared, section editing |
| Accessibility / screen readers | Multilingual v2 or Flash v2.5 | Stability over expressiveness |
GDPR and AI-Act compliance
Data residency
ElevenLabs offers EU data residency and DPA contracts for Enterprise customers. Standard- and Free-tier usage should not be used for sensitive content – clarify the current contract scope with the ElevenLabs Enterprise team.
EU AI Act and AI-generated speech
- From August 2026, synthetic speech output requires labelling under the EU AI Act
- ElevenLabs supports AI speech disclosure via metadata
- For deepfake risk: Voice Cloning for the professional variant requires KYC verification
- Recommendation: internal policy for labelling ElevenLabs audio in customer and employee communications
Copyright and personality rights
- Voice cloning only with documented consent of the cloned person
- For brand voices: contractual agreements with voice talent are mandatory
- Licence durability: When a relationship with a voice talent ends, clarify whether existing cloned voices may continue to be used
Integration into enterprise workflows
- REST API with comprehensive streaming options
- WebSocket streams for real-time conversations (Flash v2.5)
- SDKs: Python, Node.js, multiple community SDKs
- Conversational AI agents: Native integration with OpenAI, Anthropic and Gemini as LLM backbone
Our recommendation
For high-quality speech output, Eleven v3 is the leading choice in 2026, for voice agents we recommend Eleven v3 Conversational, and for commercially usable AI music, Music v2 is currently the lowest-risk option thanks to its licensed training data. For GDPR-critical applications we recommend the Enterprise tier with DPA and EU data residency and a clear internal policy for audio labelling.
As alternatives we evaluate OpenAI gpt-4o-mini-tts (well integrated with the OpenAI stack) or Cartesia Sonic (very low latency, state-space models). Contact us for advice on the right audio model strategy.