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
update
游雁abf5af40e934
diff --git a/funasr/__init__.py b/funasr/__init__.pyindex a5011bf79..950c18e80 100644--- a/funasr/__init__.py+++ b/funasr/__init__.py@@ -31,4 +31,6 @@ def import_submodules(package, recursive=True):import_submodules(__name__)from funasr.auto.auto_model import AutoModel-from funasr.auto.auto_frontend import AutoFrontend\ No newline at end of file+from funasr.auto.auto_frontend import AutoFrontend++os.environ["HYDRA_FULL_ERROR"] = 1\ No newline at end of filediff --git a/funasr/utils/load_utils.py b/funasr/utils/load_utils.pyindex 6f12f55a5..8ff7115a1 100644--- a/funasr/utils/load_utils.py+++ b/funasr/utils/load_utils.py@@ -51,13 +51,20 @@ def load_audio_text_image_video(data_or_path_or_list, fs: int = 16000, audio_fs:if isinstance(data_or_path_or_list, str) and os.path.exists(data_or_path_or_list): # local fileif data_type is None or data_type == "sound":- if use_ffmpeg:- data_or_path_or_list = _load_audio_ffmpeg(data_or_path_or_list, sr=fs)- data_or_path_or_list = torch.from_numpy(data_or_path_or_list).squeeze() # [n_samples,]- else:+ # if use_ffmpeg:+ # data_or_path_or_list = _load_audio_ffmpeg(data_or_path_or_list, sr=fs)+ # data_or_path_or_list = torch.from_numpy(data_or_path_or_list).squeeze() # [n_samples,]+ # else:+ # data_or_path_or_list, audio_fs = torchaudio.load(data_or_path_or_list)+ # if kwargs.get("reduce_channels", True):+ # data_or_path_or_list = data_or_path_or_list.mean(0)+ try:data_or_path_or_list, audio_fs = torchaudio.load(data_or_path_or_list)if kwargs.get("reduce_channels", True):data_or_path_or_list = data_or_path_or_list.mean(0)+ except:+ data_or_path_or_list = _load_audio_ffmpeg(data_or_path_or_list, sr=fs)+ data_or_path_or_list = torch.from_numpy(data_or_path_or_list).squeeze() # [n_samples,]elif data_type == "text" and tokenizer is not None:data_or_path_or_list = tokenizer.encode(data_or_path_or_list)elif data_type == "image": # undo