We're Hosting Orpheus– Real-Time, Emotional AI Voice for $1/HR

6 6 1 year ago

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.

vrfinal• 1 year ago

Are you going to be hosting the multi-lingual?

bblumenfeld102• 1 year ago

Yup, we will be working on this in the next week or two.

Sean-Der• 1 year ago

What is the latest/greatest when it comes to voice models? Orpheus is what people have the most success with?

I only concentrate on the WebRTC side, not aware of all the offerings.

jackndwyer• 1 year ago

Orpheus is the first "good" open-source model. The big players thus far have all be research labs (Cartesia, OpenAI, Play, Eleven), but with Orpheus and Sesame (another one, not quite production ready), the market is changing

keepinitneil• 1 year ago

Orpheus is good. It's a very simple architecture, basically just llama3b trained to produce audio (SNAC) tokens instead of text tokens. This is nice because it benefits from the Llama open-source ecosystem which is very active. It being LLM based is also nice because it streams output and with some engineering effort supports continuation via prompt-filling (so streaming input can work).

The other one people have been excited about is Sesame. Both are very expressive and have much more room for optimization (both on the inference efficiency side quants, finetunes, distillation, etc) and on the quality side (throw more data at it). Exciting times!

We're Hosting Orpheus– Real-Time, Emotional AI Voice for $1/HR

Related Stories