luyongqiang/Qwen-TTS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

first commit

Browse Source
This commit is contained in:
BorderArea01
2026-04-24 19:34:41 +08:00
commit 4899d56d9a
31 changed files with 11866 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

192
examples/test_model_12hz_base.py Normal file
View File

@@ -0,0 +1,192 @@
# coding=utf-8
# Copyright 2026 The Alibaba Qwen team.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import time
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
def ensure_dir(d: str):
os.makedirs(d, exist_ok=True)
def run_case(tts: Qwen3TTSModel, out_dir: str, case_name: str, call_fn):
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = call_fn()
torch.cuda.synchronize()
t1 = time.time()
print(f"[{case_name}] time: {t1 - t0:.3f}s, n_wavs={len(wavs)}, sr={sr}")
for i, w in enumerate(wavs):
sf.write(os.path.join(out_dir, f"{case_name}_{i}.wav"), w, sr)
def main():
device = "cuda:0"
MODEL_PATH = "Qwen/Qwen3-TTS-12Hz-1.7B-Base/"
OUT_DIR = "qwen3_tts_test_voice_clone_output_wav"
ensure_dir(OUT_DIR)
tts = Qwen3TTSModel.from_pretrained(
MODEL_PATH,
device_map=device,
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
# Reference audio(s)
ref_audio_path_1 = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/clone_2.wav"
ref_audio_path_2 = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/clone_1.wav"
ref_audio_single = ref_audio_path_1
ref_audio_batch = [ref_audio_path_1, ref_audio_path_2]
ref_text_single = "Okay. Yeah. I resent you. I love you. I respect you. But you know what? You blew it! And thanks to you."
ref_text_batch = [
"Okay. Yeah. I resent you. I love you. I respect you. But you know what? You blew it! And thanks to you.",
"甚至出现交易几乎停滞的情况。",
]
# Synthesis targets
syn_text_single = "Good one. Okay, fine, I'm just gonna leave this sock monkey here. Goodbye."
syn_lang_single = "Auto"
syn_text_batch = [
"Good one. Okay, fine, I'm just gonna leave this sock monkey here. Goodbye.",
"其实我真的有发现，我是一个特别善于观察别人情绪的人。",
]
syn_lang_batch = ["Chinese", "English"]
common_gen_kwargs = dict(
max_new_tokens=2048,
do_sample=True,
top_k=50,
top_p=1.0,
temperature=0.9,
repetition_penalty=1.05,
subtalker_dosample=True,
subtalker_top_k=50,
subtalker_top_p=1.0,
subtalker_temperature=0.9,
)
for xvec_only in [False, True]:
mode_tag = "xvec_only" if xvec_only else "icl"
# Case 1: prompt single + synth single, direct
run_case(
tts, OUT_DIR, f"case1_promptSingle_synSingle_direct_{mode_tag}",
lambda: tts.generate_voice_clone(
text=syn_text_single,
language=syn_lang_single,
ref_audio=ref_audio_single,
ref_text=ref_text_single,
x_vector_only_mode=xvec_only,
**common_gen_kwargs,
),
)
# Case 1b: prompt single + synth single, via create_voice_clone_prompt
def _case1b():
prompt_items = tts.create_voice_clone_prompt(
ref_audio=ref_audio_single,
ref_text=ref_text_single,
x_vector_only_mode=xvec_only,
)
return tts.generate_voice_clone(
text=syn_text_single,
language=syn_lang_single,
voice_clone_prompt=prompt_items,
**common_gen_kwargs,
)
run_case(
tts, OUT_DIR, f"case1_promptSingle_synSingle_promptThenGen_{mode_tag}",
_case1b,
)
# Case 2: prompt single + synth batch, direct
run_case(
tts, OUT_DIR, f"case2_promptSingle_synBatch_direct_{mode_tag}",
lambda: tts.generate_voice_clone(
text=syn_text_batch,
language=syn_lang_batch,
ref_audio=ref_audio_single,
ref_text=ref_text_single,
x_vector_only_mode=xvec_only,
**common_gen_kwargs,
),
)
# Case 2b: prompt single + synth batch, via create_voice_clone_prompt
def _case2b():
prompt_items = tts.create_voice_clone_prompt(
ref_audio=ref_audio_single,
ref_text=ref_text_single,
x_vector_only_mode=xvec_only,
)
return tts.generate_voice_clone(
text=syn_text_batch,
language=syn_lang_batch,
voice_clone_prompt=prompt_items,
**common_gen_kwargs,
)
run_case(
tts, OUT_DIR, f"case2_promptSingle_synBatch_promptThenGen_{mode_tag}",
_case2b,
)
# Case 3: prompt batch + synth batch, direct
run_case(
tts, OUT_DIR, f"case3_promptBatch_synBatch_direct_{mode_tag}",
lambda: tts.generate_voice_clone(
text=syn_text_batch,
language=syn_lang_batch,
ref_audio=ref_audio_batch,
ref_text=ref_text_batch,
x_vector_only_mode=[xvec_only, xvec_only],
**common_gen_kwargs,
),
)
# Case 3b: prompt batch + synth batch, via create_voice_clone_prompt
def _case3b():
prompt_items = tts.create_voice_clone_prompt(
ref_audio=ref_audio_batch,
ref_text=ref_text_batch,
x_vector_only_mode=[xvec_only, xvec_only],
)
return tts.generate_voice_clone(
text=syn_text_batch,
language=syn_lang_batch,
voice_clone_prompt=prompt_items,
**common_gen_kwargs,
)
run_case(
tts, OUT_DIR, f"case3_promptBatch_synBatch_promptThenGen_{mode_tag}",
_case3b,
)
if __name__ == "__main__":
main()

77
examples/test_model_12hz_custom_voice.py Normal file
View File

@@ -0,0 +1,77 @@
# coding=utf-8
# Copyright 2026 The Alibaba Qwen team.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import time
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
def main():
device = "cuda:0"
MODEL_PATH = "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice/"
tts = Qwen3TTSModel.from_pretrained(
MODEL_PATH,
device_map=device,
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
# -------- Single (with instruct) --------
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = tts.generate_custom_voice(
text="其实我真的有发现，我是一个特别善于观察别人情绪的人。",
language="Chinese",
speaker="Vivian",
instruct="用特别愤怒的语气说",
)
torch.cuda.synchronize()
t1 = time.time()
print(f"[CustomVoice Single] time: {t1 - t0:.3f}s")
sf.write("qwen3_tts_test_custom_single.wav", wavs[0], sr)
# -------- Batch (some empty instruct) --------
texts = ["其实我真的有发现，我是一个特别善于观察别人情绪的人。", "She said she would be here by noon."]
languages = ["Chinese", "English"]
speakers = ["Vivian", "Ryan"]
instructs = ["", "Very happy."]
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = tts.generate_custom_voice(
text=texts,
language=languages,
speaker=speakers,
instruct=instructs,
max_new_tokens=2048,
)
torch.cuda.synchronize()
t1 = time.time()
print(f"[CustomVoice Batch] time: {t1 - t0:.3f}s")
for i, w in enumerate(wavs):
sf.write(f"qwen3_tts_test_custom_batch_{i}.wav", w, sr)
if __name__ == "__main__":
main()

80
examples/test_model_12hz_voice_design.py Normal file
View File

@@ -0,0 +1,80 @@
# coding=utf-8
# Copyright 2026 The Alibaba Qwen team.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import time
import torch
import soundfile as sf
from qwen_tts import Qwen3TTSModel
def main():
device = "cuda:0"
MODEL_PATH = "Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign/"
tts = Qwen3TTSModel.from_pretrained(
MODEL_PATH,
device_map=device,
dtype=torch.bfloat16,
attn_implementation="flash_attention_2",
)
# -------- Single --------
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = tts.generate_voice_design(
text="哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
language="Chinese",
instruct="体现撒娇稚嫩的萝莉女声，音调偏高且起伏明显，营造出黏人、做作又刻意卖萌的听觉效果。",
)
torch.cuda.synchronize()
t1 = time.time()
print(f"[VoiceDesign Single] time: {t1 - t0:.3f}s")
sf.write("qwen3_tts_test_voice_design_single.wav", wavs[0], sr)
# -------- Batch --------
texts = [
"哥哥，你回来啦，人家等了你好久好久了，要抱抱！",
"It's in the top drawer... wait, it's empty? No way, that's impossible! I'm sure I put it there!"
]
languages = ["Chinese", "English"]
instructs = [
"体现撒娇稚嫩的萝莉女声，音调偏高且起伏明显，营造出黏人、做作又刻意卖萌的听觉效果。",
"Speak in an incredulous tone, but with a hint of panic beginning to creep into your voice."
]
torch.cuda.synchronize()
t0 = time.time()
wavs, sr = tts.generate_voice_design(
text=texts,
language=languages,
instruct=instructs,
max_new_tokens=2048,
)
torch.cuda.synchronize()
t1 = time.time()
print(f"[VoiceDesign Batch] time: {t1 - t0:.3f}s")
for i, w in enumerate(wavs):
sf.write(f"qwen3_tts_test_voice_design_batch_{i}.wav", w, sr)
if __name__ == "__main__":
main()

65
examples/test_tokenizer_12hz.py Normal file
View File

@@ -0,0 +1,65 @@
# coding=utf-8
# Copyright 2026 The Alibaba Qwen team.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import io
import requests
import soundfile as sf
from qwen_tts import Qwen3TTSTokenizer
audio_1 = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/tokenizer_demo_1.wav"
audio_2 = "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-TTS-Repo/tokenizer_demo_2.wav"
# -------- Single input: wav path --------
tokenizer_12hz = Qwen3TTSTokenizer.from_pretrained(
"Qwen/Qwen3-TTS-Tokenizer-12Hz",
device_map="cuda:0",
)
enc1 = tokenizer_12hz.encode(audio_1)
wavs1, out_sr1 = tokenizer_12hz.decode(enc1)
sf.write("decoded_single_12hz.wav", wavs1[0], out_sr1)
# -------- Batch input: wav path list --------
enc2 = tokenizer_12hz.encode([audio_1, audio_2])
wavs2, out_sr2 = tokenizer_12hz.decode(enc2)
for i, w in enumerate(wavs2):
sf.write(f"decoded_batch_12hz_{i}.wav", w, out_sr2)
# -------- Decode input as dict (12hz) --------
# Take the first sample codes and pass as a dict.
dict_input_12hz = {"audio_codes": enc2.audio_codes[0]} # torch.Tensor
wavs_d1, out_sr_d1 = tokenizer_12hz.decode(dict_input_12hz)
sf.write("decoded_dict_12hz.wav", wavs_d1[0], out_sr_d1)
# -------- Decode input as list[dict] (12hz) --------
list_dict_input_12hz = [{"audio_codes": c} for c in enc2.audio_codes] # list of torch.Tensor
wavs_d2, out_sr_d2 = tokenizer_12hz.decode(list_dict_input_12hz)
for i, w in enumerate(wavs_d2):
sf.write(f"decoded_listdict_12hz_{i}.wav", w, out_sr_d2)
# -------- Decode input as list[dict] with numpy (12hz) --------
# Convert codes to numpy to simulate "serialized" payload.
list_dict_numpy_12hz = [{"audio_codes": c.cpu().numpy()} for c in enc2.audio_codes]
wavs_d3, out_sr_d3 = tokenizer_12hz.decode(list_dict_numpy_12hz)
for i, w in enumerate(wavs_d3):
sf.write(f"decoded_listdict_numpy_12hz_{i}.wav", w, out_sr_d3)
# -------- Numpy input (must pass sr) --------
data = requests.get(audio_2, timeout=30).content
y, sr = sf.read(io.BytesIO(data))
enc3 = tokenizer_12hz.encode(y, sr=sr)
wavs3, out_sr3 = tokenizer_12hz.decode(enc3)
sf.write("decoded_numpy_12hz.wav", wavs3[0], out_sr3)