minssam.
Published on

ElevenLabs Major Update: SFX v3 for Everyone and Real-Time Voice Agents Are Here

The world of AI-generated sound just got more complete.

ElevenLabs is best known for its text-to-speech technology, but by 2026 it had grown far beyond basic voice synthesis. The November 2, 2026 update reflects that growth clearly.

SFX v3: The Most Advanced Sound Effects Model Yet

The first headline from this update is SFX v3. Its model identifier is eleven_text_to_sound_v3, and it is now the most capable sound effects model in the ElevenLabs catalog.

SFX v3 goes beyond generating simple one-shot sounds. It is now the default model in ElevenCreative''s Sound Effects, Studio, and Flows menus — meaning users get it automatically without changing any settings.

The practical applications are broad: video editing, podcast production, game development, educational content, and any workflow that calls for custom sound design.

v4 Voice Clone Fine-Tuning: Open to All Workspaces

The second highlight is that Professional Voice Clone (PVC) fine-tuning on Eleven v4 and v4 Turbo is now open to every workspace — not just a select group of users.

Voice clone owners can now go into My Voices and retrain their existing clones on the v4 engine for more natural, expressive output. One practical improvement: when sending requests to a v4-based clone, you no longer need to pass the clone_strength parameter on every call. The system falls back to the value saved in the voice''s own settings automatically.

ElevenAgents and Eleven Realtime: Real-Time Voice Agents, Fully Launched

The third major change is in ElevenAgents — the platform formerly called Conversational AI, now rebranded. The key improvement here is that Eleven Realtime is now available for voice agents in all workspaces. The rollout started October 19 and is now complete.

Agent builders can choose between Eleven Realtime v1 Mini (optimized for speed) and Eleven Realtime v1 Max (optimized for quality) as the voice mode.

Noisy-environment detection can now be configured through the API and CLI — a useful feature for call center agents or public-space conversational systems.

Expanded AI Model Support for Agents

This update also expands which AI models voice agents can use as their underlying brain: gemini-4-argon, claude-opus-4-6, and ElevenLabs-hosted nemotron-35-super-vl-nvfp4 are all now options. This means more control over the intelligence behind your voice agent.

Practical Tips

  • Content creators: Use SFX v3 to generate natural-sounding effects for YouTube, Reels, or TikTok without separate audio software.
  • Brand voice: Clone and fine-tune a unique brand voice on Eleven v4 for consistent, high-quality TTS across all materials.
  • EdTech developers: Build real-time speaking practice agents with ElevenAgents + Eleven Realtime — the noisy-environment detection makes it realistic.
  • Developers: Evaluation criteria can now be saved as reusable analysis items, making agent quality management more systematic.

Sources:

ElevenLabs Major Update: SFX v3 for Everyone and Real-Time Voice Agents Are Here | MINSSAM.COM