Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.
Python asraudiochineseemotion-recognitionfunasrmcp-servermultilingual-asropenai-compatible-apiparaformerpunctuationpytorchreal-time-asrspeaker-diarizationspeech-recognitionspeech-to-textstreaming-asrtranscriptionvllmvoice-activity-detectionwhisper-alternative
sensevoice
游雁8b74979bd38c
diff --git a/examples/industrial_data_pretraining/sense_voice/demo.py b/examples/industrial_data_pretraining/sense_voice/demo.pyindex 1800d7591..f10e54b45 100644--- a/examples/industrial_data_pretraining/sense_voice/demo.py+++ b/examples/industrial_data_pretraining/sense_voice/demo.py@@ -3,26 +3,82 @@# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.# MIT License (https://opensource.org/licenses/MIT)-import sys+from funasr import AutoModelfrom funasr.utils.postprocess_utils import rich_transcription_postprocessmodel_dir = "iic/SenseVoiceSmall"-input_file = (- "https://isv-data.oss-cn-hangzhou.aliyuncs.com/ics/MaaS/ASR/test_audio/asr_example_zh.wav"-)+model = AutoModel(model=model_dir,+ vad_model="fsmn-vad",+ vad_kwargs={"max_single_segment_time": 30000},+ device="cpu",+)++# en+res = model.generate(+ input=f"{model.model_path}/example/en.mp3",+ cache={},+ language="auto", # "zn", "en", "yue", "ja", "ko", "nospeech"+ use_itn=True,+ batch_size_s=60,+ merge_vad=True, #+ merge_length_s=15,+)+text = rich_transcription_postprocess(res[0]["text"])+print(text)++# zh+res = model.generate(+ input=f"{model.model_path}/example/zh.mp3",+ cache={},+ language="auto", # "zn", "en", "yue", "ja", "ko", "nospeech"+ use_itn=True,+ batch_size_s=60,+ merge_vad=True, #+ merge_length_s=15,)+text = rich_transcription_postprocess(res[0]["text"])+print(text)+# yueres = model.generate(- input=input_file,+ input=f"{model.model_path}/example/yue.mp3",cache={},language="auto", # "zn", "en", "yue", "ja", "ko", "nospeech"- use_itn=False,+ use_itn=True,+ batch_size_s=60,+ merge_vad=True, #+ merge_length_s=15,)+text = rich_transcription_postprocess(res[0]["text"])+print(text)+# ja+res = model.generate(+ input=f"{model.model_path}/example/ja.mp3",+ cache={},+ language="auto", # "zn", "en", "yue", "ja", "ko", "nospeech"+ use_itn=True,+ batch_size_s=60,+ merge_vad=True, #+ merge_length_s=15,+)text = rich_transcription_postprocess(res[0]["text"])+print(text)++# ko+res = model.generate(+ input=f"{model.model_path}/example/ko.mp3",+ cache={},+ language="auto", # "zn", "en", "yue", "ja", "ko", "nospeech"+ use_itn=True,+ batch_size_s=60,+ merge_vad=True, #+ merge_length_s=15,+)+text = rich_transcription_postprocess(res[0]["text"])print(text)diff --git a/funasr/auto/auto_model.py b/funasr/auto/auto_model.pyindex 1c2d2e48f..a82f6ed75 100644--- a/funasr/auto/auto_model.py+++ b/funasr/auto/auto_model.py@@ -339,7 +339,9 @@ class AutoModel:# FIX(gcf): concat the vad clips for sense vocie model for better aedif kwargs.get("merge_vad", False):for i in range(len(res)):- res[i]["value"] = merge_vad(res[i]["value"], kwargs.get("merge_length", 15000))+ res[i]["value"] = merge_vad(+ res[i]["value"], kwargs.get("merge_length_s", 15) * 1000+ )# step.2 compute asr modelmodel = self.model@@ -380,6 +382,9 @@ class AutoModel:if len(sorted_data) > 0 and len(sorted_data[0]) > 0:batch_size = max(batch_size, sorted_data[0][0][1] - sorted_data[0][0][0])+ if kwargs["device"] == "cpu":+ batch_size = 0+beg_idx = 0beg_asr_total = time.time()time_speech_total_per_sample = speech_lengths / 16000