Supported Languages and Dialects in Qwen3-ASR

import torch
from qwen_asr import Qwen3ASRModel

model = Qwen3ASRModel.from_pretrained(
    "Qwen/Qwen3-ASR-1.7B",
    dtype=torch.bfloat16,
    device_map="cuda:0",
)

# Single audio — force English
results = model.transcribe(
    audio="path/to/english_audio.wav",
    language="English",
)
print(results[0].text)

# Batch — specify a language per item
results = model.transcribe(
    audio=[
        "path/to/chinese_audio.wav",
        "path/to/french_audio.wav",
    ],
    language=["Chinese", "French"],
)
for r in results:
    print(r.language, r.text)

import torch
from qwen_asr import Qwen3ASRModel

model = Qwen3ASRModel.from_pretrained(
    "Qwen/Qwen3-ASR-1.7B",
    dtype=torch.bfloat16,
    device_map="cuda:0",
)

# Auto-detect language
results = model.transcribe(
    audio="path/to/unknown_language.wav",
    language=None,  # None triggers automatic language detection
)

print(results[0].language)   # e.g., "Japanese"
print(results[0].text)

import torch
from qwen_asr import Qwen3ForcedAligner

aligner = Qwen3ForcedAligner.from_pretrained(
    "Qwen/Qwen3-ForcedAligner-0.6B",
    dtype=torch.bfloat16,
    device_map="cuda:0",
)

results = aligner.align(
    audio="path/to/audio.wav",
    text="甚至出现交易几乎停滞的情况。",
    language="Chinese",   # Must be one of the 11 aligner languages
)

for token in results[0]:
    print(f"{token.text}  {token.start_time:.3f}s – {token.end_time:.3f}s")