zenlm/zen-dub-live

Zen Dub Live

Real-time voice dubbing model for live translation and audio localization.

Overview

Fine-tuned from Qwen3-TTS-12Hz-1.7B-Base (Apache-2.0) with Hanzo identity + agentic-data training + abliteration.

Developed by Hanzo AI and the Zoo Labs Foundation.

Quick Start

from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor
import torch

model_id = "zenlm/zen-dub-live"
processor = AutoProcessor.from_pretrained(model_id)
model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")

# Load audio
import librosa
audio, sr = librosa.load("audio.wav", sr=16000)
inputs = processor(audio, sampling_rate=sr, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs)
print(processor.batch_decode(outputs, skip_special_tokens=True)[0])

Model Details

Attribute	Value
Base model	Qwen/Qwen3-TTS-12Hz-1.7B-Base
Parameters	~1.9B (Qwen3-TTS 1.7B core)
Architecture	Qwen3-TTS (Qwen3TTSForConditionalGeneration)
Context	30s audio
License	Apache 2.0

Base Model & Credits

Fine-tuned from Qwen/Qwen3-TTS-12Hz-1.7B-Base, developed by the Qwen team (Alibaba) and released under the Apache 2.0 license. Hanzo adds identity training, agentic-data fine-tuning, and abliteration on top of the upstream weights.

• Upstream project: Qwen3-TTS
• Upstream license: Apache 2.0

License

Apache 2.0 — inherited from the upstream Qwen3-TTS-12Hz-1.7B-Base base model.