Eleven Labs Alternatives

Direct alternatives to ElevenLabs

ElevenLabs covers text-to-speech, voice cloning, dubbing, speech-to-text and music behind one API. Teams usually look elsewhere for one piece of that: lower latency inside a voice agent, weights they can run themselves, or hosting in a specific jurisdiction. The list below is grouped by that reason.

Real-time speech for voice agents

• Cartesia: Sonic text-to-speech built on state space models, with 44 languages on Sonic 3.6 as of September 2026. Cartesia offers cloud, on-premise and on-device deployment, which matters when an agent cannot call out to a public API.
• Rime AI: TTS API for conversational agents, listing 600+ voices and 50+ languages as of September 2026. It runs as a cloud API, inside a VPC or fully on-premise.
• SpeechifyAI: the developer API for Speechify's Simba models, with zero-shot voice cloning from a short clip. Speechify publishes a p50 first-byte latency of 56 ms on its own production path (September 2026).

Open weights to self-host

• VoxCPM: OpenBMB's text-to-speech model under Apache 2.0, with voice cloning and 30 languages. The 2B-parameter model runs in roughly 8GB of VRAM with no per-character bill.
• Fish Audio: hosted TTS and voice cloning, with the Fish Speech code and weights on GitHub under the Fish Audio Research License. Check those terms before commercial use.

Hosting in Europe

• KugelAudio: Berlin TTS provider running its own models in the EU, with voice cloning included on every model and billing per generated audio minute. Kugel 3 covers 26 languages.
• Samtal: Swedish-hosted TTS, speech-to-text and voice cloning behind an ElevenLabs-compatible API. Existing ElevenLabs SDK code works by changing the base URL.