Models
zen-live
Real-time bidirectional speech translation with ultra-low latency.
zen-live
Real-Time Translation
A real-time bidirectional translation model that transcribes, translates, and speaks in a single streaming pipeline. Enables live cross-language conversations with minimal latency.
This model is coming soon. Join the waitlist at hanzo.chat.
Specifications
| Property | Value |
|---|---|
| Model ID | zen-live |
| Architecture | Streaming Multimodal Transformer |
| Latency | < 500ms end-to-end |
| Languages | 50+ |
| Status | Coming Soon |
| HuggingFace | -- |
Capabilities
- Real-time speech-to-speech translation
- Bidirectional conversation mode
- Sub-500ms end-to-end latency
- Tone and intent preservation
- Speaker voice adaptation
- Multi-party conference translation
Usage
from hanzoai import Hanzo
client = Hanzo(api_key="hk-your-api-key")
# Coming soon -- real-time streaming API
session = client.realtime.create(
model="zen-live",
source_language="en",
target_language="ja",
)
# Stream audio in, get translated audio out
for translated_audio in session.stream(microphone_input):
speaker.play(translated_audio)See Also
- zen-translator -- Text translation (100+ languages)
- zen-scribe -- Speech-to-text transcription
- zen-dub-live -- Real-time voice synthesis