We just launched hosted Orpheus on [Gabber.dev](
https://gabber.dev/)—an LLM-agnostic voice API for building real-time, emotionally expressive AI agents.
We’re hosting [Orpheus TTS](https://github.com/canopy-ai/orpheus-tts) in production, which gives you:
- ~200ms TTS streaming latency
- Expressive voices that sound like humans
- Emotion tags like <laugh>, <sigh>, <cry>
- Significantly lower cost (~$1/hr) compared to ElevenLabs or [Play.ht](http://Play.ht) ($5+/hr)
- Integrate with WebSocket API or Realtime WebRTC SDK
Gabber also handles:
- Memory (long-term + session-based)
- LLM orchestration (use your own or ours)
- Tool calling
With this release, real-time AI voice is finally fast, affordable, and expressive enough to unlock use cases that just weren’t viable before. With cheaper, easier access to high-quality voice, you can now build what you actually want—whether it’s an AI companion, a voice-native coach, a video game NPC, or something entirely your own.
If you're curious about how we’re hosting Orpheus at scale or want to talk memory, GPU infra, SDKs—happy to share.
We’re hosting [Orpheus TTS](https://github.com/canopy-ai/orpheus-tts) in production, which gives you:
- ~200ms TTS streaming latency
- Expressive voices that sound like humans
- Emotion tags like <laugh>, <sigh>, <cry>
- Significantly lower cost (~$1/hr) compared to ElevenLabs or [Play.ht](http://Play.ht) ($5+/hr)
- Integrate with WebSocket API or Realtime WebRTC SDK
Gabber also handles:
- Memory (long-term + session-based)
- LLM orchestration (use your own or ours)
- Tool calling
With this release, real-time AI voice is finally fast, affordable, and expressive enough to unlock use cases that just weren’t viable before. With cheaper, easier access to high-quality voice, you can now build what you actually want—whether it’s an AI companion, a voice-native coach, a video game NPC, or something entirely your own.
If you're curious about how we’re hosting Orpheus at scale or want to talk memory, GPU infra, SDKs—happy to share.