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
vad conf
游雁d29f201e3201
diff --git a/examples/industrial_data_pretraining/emotion2vec/demo.py b/examples/industrial_data_pretraining/emotion2vec/demo.pyindex b274bd966..940b9f932 100644--- a/examples/industrial_data_pretraining/emotion2vec/demo.py+++ b/examples/industrial_data_pretraining/emotion2vec/demo.py@@ -9,7 +9,7 @@ from funasr import AutoModelmodel = AutoModel(model="iic/emotion2vec_base_finetuned", model_revision="v2.0.4",# vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",# vad_model_revision="v2.0.4",- # vad_kwargs={"max_single_segment_time": 10},+ # vad_kwargs={"max_single_segment_time": 1000},)wav_file = f"{model.model_path}/example/test.wav"diff --git a/examples/industrial_data_pretraining/paraformer/demo.py b/examples/industrial_data_pretraining/paraformer/demo.pyindex 499791fa2..a39a52673 100644--- a/examples/industrial_data_pretraining/paraformer/demo.py+++ b/examples/industrial_data_pretraining/paraformer/demo.py@@ -9,7 +9,7 @@ model = AutoModel(model="iic/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-model_revision="v2.0.4",vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",vad_model_revision="v2.0.4",- vad_kwargs={"max_single_segment_time": 60},+ vad_kwargs={"max_single_segment_time": 60000},punc_model="iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch",punc_model_revision="v2.0.4",# spk_model="iic/speech_campplus_sv_zh-cn_16k-common",diff --git a/examples/industrial_data_pretraining/whisper/demo.py b/examples/industrial_data_pretraining/whisper/demo.pyindex a073f68f3..e1e1aad29 100644--- a/examples/industrial_data_pretraining/whisper/demo.py+++ b/examples/industrial_data_pretraining/whisper/demo.py@@ -10,7 +10,7 @@ from funasr import AutoModelmodel = AutoModel(model="iic/Whisper-large-v3",model_revision="v2.0.5",vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",- vad_kwargs={"max_single_segment_time": 30},+ vad_kwargs={"max_single_segment_time": 30000},)res = model.generate(diff --git a/examples/industrial_data_pretraining/whisper/demo_from_openai.py b/examples/industrial_data_pretraining/whisper/demo_from_openai.pyindex 9a2276445..a257bc820 100644--- a/examples/industrial_data_pretraining/whisper/demo_from_openai.py+++ b/examples/industrial_data_pretraining/whisper/demo_from_openai.py@@ -12,7 +12,7 @@ from funasr import AutoModel# model = AutoModel(model="Whisper-large-v2", hub="openai")model = AutoModel(model="Whisper-large-v3",vad_model="iic/speech_fsmn_vad_zh-cn-16k-common-pytorch",- vad_kwargs={"max_single_segment_time": 30},+ vad_kwargs={"max_single_segment_time": 30000},hub="openai",)diff --git a/funasr/auto/auto_model.py b/funasr/auto/auto_model.pyindex bba44e792..bd786d0b6 100644--- a/funasr/auto/auto_model.py+++ b/funasr/auto/auto_model.py@@ -312,7 +312,8 @@ class AutoModel:key = res[i]["key"]vadsegments = res[i]["value"]input_i = data_list[i]- speech = load_audio_text_image_video(input_i, fs=kwargs["frontend"].fs, audio_fs=kwargs.get("fs", 16000))+ fs = kwargs["frontend"].fs if hasattr(kwargs["frontend"], "fs") else 16000+ speech = load_audio_text_image_video(input_i, fs=fs, audio_fs=kwargs.get("fs", 16000))speech_lengths = len(speech)n = len(vadsegments)data_with_index = [(vadsegments[i], i) for i in range(n)]