add recordings and tests

lib/utils.py

@@ -38,7 +38,7 @@ def cmd(command: str, check=True, capture_output=False) -> CompletedProcess:
     return ret
 
 def clean_text_for_comparison_zh(text):
-    symbol_pattern = "[ ，。、!?\n]"
+    symbol_pattern = "[ ，。、!?:：‘’-《》！？；,\n]"
     to = ""
     return re.sub(symbol_pattern, to, text).lower()
 

scripts/asr_utils.py

@@ -0,0 +1,81 @@
+import os
+import csv
+import wave
+import re
+
+def add_text_index():
+    text_file = '/Users/jeqin/work/code/TestTranslator/tests/test_data/text/test_asr_zh.txt'
+    index = 1
+    with open(text_file, encoding='utf-8') as f:
+        for line in f:
+            line = line.strip()
+            # print(line)
+            if not line:
+                continue
+            if line.startswith('#'):
+                # print(line)
+                continue
+            line = f"{index}. {line}"
+            print(line)
+            index += 1
+
+def get_lines_with_index(filepath):
+    with open(filepath, encoding='utf-8') as f:
+        for line in f:
+            line = line.strip()
+            m = re.match(r'^(\d+)\.\s*(.*)', line)
+            if m:
+                yield m.group(1), m.group(2)
+
+def get_wav_length(wav_path):
+    try:
+        with wave.open(wav_path, 'rb') as wf:
+            frames = wf.getnframes()
+            rate = wf.getframerate()
+            duration = frames / float(rate)
+            return duration
+    except Exception as e:
+        print(f"Error reading {wav_path}: {e}")
+        return 0
+
+def write_csv(rows, output_csv):
+    with open(output_csv, 'w', newline='', encoding='utf-8') as csvfile:
+        writer = csv.writer(csvfile)
+        writer.writerow(['序号', '文本', '音频长度(秒)'])
+        writer.writerows(rows)
+
+def print_text_and_audio_length():
+    text_file = '/Users/jeqin/work/code/TestTranslator/tests/test_data/text/test_asr_zh_with_index.txt'
+    audio_folder = '/Users/jeqin/work/code/TestTranslator/tests/test_data/recordings'
+    output_csv = '/Users/jeqin/work/code/TestTranslator/scripts/csv/text_audio_length.csv'
+    rows = []
+    for idx, text in get_lines_with_index(text_file):
+        # print(idx)
+        # print(text)
+        audio_path = os.path.join(audio_folder, f"{idx}.wav")
+        audio_length = get_wav_length(audio_path)
+        audio_length = round(audio_length, 2) if audio_length is not None else None
+        # print(audio_length)
+        rows.append([idx, text, round(audio_length,2)])
+    write_csv(rows, output_csv)
+
+def get_text_distance(text1, text2):
+    from lib.utils import run_textdistance, clean_text_for_comparison_zh, highlight_diff
+    text1_clean = clean_text_for_comparison_zh(text1)
+    text2_clean = clean_text_for_comparison_zh(text2)
+    d, nd = run_textdistance(text1_clean, text2_clean)
+    diff = highlight_diff(text1_clean, text2_clean, spliter="")
+    return d, nd, diff
+
+def get_origin_text_dict():
+    text_file = '/Users/jeqin/work/code/TestTranslator/tests/test_data/text/test_asr_zh_with_index.txt'
+    text_dict = {}
+    for idx, text in get_lines_with_index(text_file):
+        text_dict[idx] = text
+    return text_dict
+
+
+if __name__ == '__main__':
+    # add_text_index()
+    # print_text_and_audio_length()
+    pass

scripts/infer_finetuned_whisper.py

@@ -112,9 +112,10 @@ def main():
                 print(f"{p.name} -> {text}; time cost: {t1-t0}")
             except Exception as e:
                 print(f"{p.name} -> 失败: {e}")
-def run():
+
+def load_model():
     model_path = "/Users/jeqin/Downloads/whisper-large-v3-turbo-finetune-0901"
-    lang = "en"
+    lang = "zh"
     t0 = time.time()
     processor = WhisperProcessor.from_pretrained(
         model_path,
@@ -133,9 +134,11 @@ def run():
     model.generation_config.language = lang.lower()
     model.generation_config.forced_decoder_ids = None
     model.eval()
-
-
     print("load model time: ", time.time() - t0)
+    return model, processor
+
+def run_test_audios():
+    model, processor = load_model()
     audios = Path("/Users/jeqin/work/code/TestTranslator/tests/test_data/test_audios/")
     rows = [["file_name", "inference_time", "inference_result"]]
     for audio in sorted(audios.glob("*en-ac1-16k/*.wav")): # *s/randomforest*.wav"
@@ -152,6 +155,27 @@ def run():
             print(f"{audio.name} -> 失败: {e}")
     save_csv("csv/fine-tune_whisper-0901.csv", rows)
 
+def run_recordings():
+    from scripts.asr_utils import get_origin_text_dict, get_text_distance
+    model, processor = load_model()
+    audios = Path("/Users/jeqin/work/code/TestTranslator/tests/test_data/recordings/")
+    rows = [["file_name", "time", "inference_result"]]
+    original = get_origin_text_dict()
+    for audio in sorted(audios.glob("*.wav"), key=lambda x: int(x.stem)):
+        print(audio)
+        try:
+            t0 = time.time()
+            text = transcribe_file(
+                str(audio), model, processor
+            )
+            t = time.time()-t0
+            print(text)
+            print("inference time:", t)
+            d, nd, diff = get_text_distance(original[audio.stem], text)
+            rows.append([audio.name, round(t, 3), text, d, round(nd,3), diff])
+        except Exception as e:
+            print(f"{audio.name} -> 失败: {e}")
+    save_csv("csv/fine-tune_whisper.csv", rows)
 if __name__ == "__main__":
     # main()
-    run()
+    run_recordings()

scripts/recorder.py

@@ -0,0 +1,56 @@
+import sounddevice as sd
+import soundfile as sf
+import sys
+import re
+
+TEXT_FILE = '/Users/jeqin/work/code/TestTranslator/tests/test_data/text/test_asr_zh_with_index.txt'
+AUDIO_FOLDER= '/Users/jeqin/work/code/TestTranslator/tests/test_data/recordings'
+SAMPLE_RATE = 16000
+CHANNELS = 1
+
+def get_lines_with_index(filepath):
+    with open(filepath, encoding='utf-8') as f:
+        for line in f:
+            line = line.strip()
+            m = re.match(r'^(\d+)\.\s*(.*)', line)
+            if m:
+                yield m.group(1), m.group(2)
+
+def record_audio(filename):
+    import numpy as np
+    def callback(indata, frames, time, status):
+        recording.append(indata.copy())
+
+    while True:
+        print("按回车开始录音...")
+        input()
+        print("正在录音，按回车结束录音，或输入 q 回车重新录音。")
+        recording = []
+        stop = False
+
+        with sd.InputStream(samplerate=SAMPLE_RATE, channels=CHANNELS, dtype='float32', callback=callback):
+            user_input = input()
+            if user_input.strip().lower() == 'q':
+                print("重新录音...")
+                stop = False
+                continue  # 跳出 with，重新录音
+            else:
+                stop = True
+
+        if stop:
+            audio_np = np.concatenate(recording, axis=0)
+            max_val = np.max(np.abs(audio_np))
+            if max_val > 0:
+                audio_np = audio_np * (0.99 / max_val)
+            sf.write(f"{AUDIO_FOLDER}/{filename}", audio_np, SAMPLE_RATE)
+            print(f"已保存: {filename}")
+            break
+
+def main():
+    for idx, text in get_lines_with_index(TEXT_FILE):
+        print(f"{idx}. {text}")
+        if int(idx)==52:
+            record_audio(f"{idx}.wav")
+
+if __name__ == '__main__':
+    main()

scripts/run_funasr.py

@@ -27,7 +27,7 @@ def main():
     print("load model: ", t1 - t0)
     audios = Path("/Users/jeqin/work/code/TestTranslator/tests/test_data/test_audios/")
     rows = [["file_name", "inference_time", "inference_result"]]
-    for audio in sorted(audios.glob("*ac1-16k/Chinese*")):
+    for audio in sorted(audios.glob("*mix/*")):
         print(audio)
         t1 = time.time()
         try:

scripts/{run_quant.py → run_funasr_quant.py}

@@ -2,13 +2,15 @@ from pathlib import Path
 import time
 import csv
 from funasr_onnx import SeacoParaformer, CT_Transformer, Fsmn_vad
+from scripts.asr_utils import get_origin_text_dict, get_text_distance
 
 def save_csv(file_path, rows):
     with open(file_path, "w", encoding="utf-8") as f:
         writer = csv.writer(f)
         writer.writerows(rows)
         print(f"write csv to {file_path}")
-def main():
+
+def load_model(quantize=True):
     model_dir = Path("/Users/jeqin/work/code/Translator/python_server/moyoyo_asr_models")
 
     asr_model_path = model_dir / 'speech_seaco_paraformer_large_asr_nat-zh-cn-16k-common-vocab8404-pytorch'
@@ -21,31 +23,52 @@ def main():
     punc_model = CT_Transformer(punc_model_path, quantize=quantize)
     t1 = time.time()
     print("load model time:", t1 - t0)
+    return vad_model, asr_model, punc_model
+
+def inference(vad_model, asr_model, punc_model, audio:Path):
+    print(audio.name)
+    t1 = time.time()
+    vad_res = vad_model(str(audio))
+    t2 = time.time()
+    # print("vad time:", t2-t1)
+    asr_res = asr_model(str(audio), hotwords="")
+    asr_text = asr_res[0]["preds"]
+    t3 = time.time()
+    # print("asr time:", t3-t2)
+    # print("asr text:", asr_text)
+    result = punc_model(asr_text)
+    text = result[0]
+    t4 = time.time()
+    # print("punc time:", t4-t3)
+    # print("punc text:", text)
+    print(text)
+    t = t4-t1
+    print("inference:", t)
+    return text, t
+
+def run_recordings():
+    quantize = True
+    vad_model, asr_model, punc_model = load_model(quantize)
+    audios = Path("/Users/jeqin/work/code/TestTranslator/tests/test_data/recordings/")
+    rows = [["file_name", "time", "inference_result"]]
+    original = get_origin_text_dict()
+    for audio in sorted(audios.glob("*.wav"), key=lambda x: int(x.stem)):
+        text, t = inference(vad_model, asr_model, punc_model, audio)
+        d, nd, diff = get_text_distance(original[audio.stem], text)
+        rows.append([audio.name, round(t, 3), text, d, round(nd,3), diff]) # f"{audio.parent.name}/{audio.name}"
+    file_name = "csv/funasr_quant.csv" if quantize else "funasr_onnx.csv"
+    save_csv(file_name, rows)
+
+def run_test_audios():
+    quantize = True
+    vad_model, asr_model, punc_model = load_model(quantize)
     audios = Path("/Users/jeqin/work/code/TestTranslator/tests/test_data/test_audios/")
-    rows = [["file_name", "inference_time", "inference_result"]]
-    for audio in sorted(audios.glob("*s/randomforest*.wav")):
-        t1 = time.time()
-        vad_res = vad_model(str(audio))
-        t2 = time.time()
-        print("vad time:", t2-t1)
-        asr_res = asr_model(str(audio), hotwords="")
-        asr_text = asr_res[0]["preds"]
-        t3 = time.time()
-        print("asr time:", t3-t2)
-        print("asr text:", asr_text)
-        result = punc_model(asr_text)
-        text = result[0]
-        t4 = time.time()
-        print("punc time:", t4-t3)
-        print("punc text:", text)
-        # print(text)
-        # vad_res = vad_model(str(audio))
-        # t5 = time.time()
-        # print("vad time:", t5 - t4)
-        t = t4-t1
-        print("inference:", t)
-        rows.append([f"{audio.parent.name}/{audio.name}", t, text])
-    file_name = "csv/quant.csv" if quantize else "run_onnx.csv"
+    rows = [["file_name", "time", "inference_result"]]
+    for audio in sorted(audios.glob("*s/zh*.wav")):
+        text, t = inference(vad_model, asr_model, punc_model, audio)
+        rows.append([f"{audio.parent.name}/{audio.name}", round(t, 3), text])
+    file_name = "csv/funasr_quant.csv" if quantize else "funasr_onnx.csv"
     save_csv(file_name, rows)
+    
 if __name__ == '__main__':
-    main()
+    run_recordings()

scripts/run_whisper.py

@@ -4,7 +4,7 @@ import time
 import csv
 
 from silero_vad.utils_vad import languages
-
+from scripts.asr_utils import get_origin_text_dict, get_text_distance
 
 def save_csv(file_path, rows):
     with open(file_path, "w", encoding="utf-8") as f:
@@ -12,7 +12,7 @@ def save_csv(file_path, rows):
         writer.writerows(rows)
         print(f"write csv to {file_path}")
 
-def main():
+def load_model():
     models_dir = Path("/Users/jeqin/work/code/Translator/python_server/moyoyo_asr_models")
     whisper_model = 'large-v3-turbo-q5_0'
     t0 = time.time()
@@ -28,18 +28,40 @@ def main():
                 no_context=True
             )
     print("load model time: ", time.time()-t0)
-    audios = Path("/Users/jeqin/work/code/TestTranslator/tests/test_data/test_audios/")
-    rows = [["file_name", "inference_time", "inference_result"]]
-    for audio in sorted(audios.glob("*-mix/randomforest*.wav")):
+    return model
+
+def run_recordings():
+    model = load_model()
+    audios = Path("/Users/jeqin/work/code/TestTranslator/tests/test_data/recordings/")
+    rows = [["file_name", "time", "inference_result"]]
+    original = get_origin_text_dict()
+    for audio in sorted(audios.glob("*.wav"), key=lambda x: int(x.stem)):
         print(audio)
         t1 = time.time()
-        output = model.transcribe(str(audio), language="zh")#, language="zh", initial_prompt="这是一段中文的会议内容。")# initial_prompt="这是一段中文的会议内容。"
+        output = model.transcribe(str(audio), language="zh", initial_prompt="以下是普通话句子，这是一段会议内容。")# initial_prompt="这是一段中文的会议内容。"
         t = time.time() - t1
         print("inference time:", t)
         text = " ".join([a.text for a in output])
         print(text)
-        rows.append([f"{audio.parent.name}/{audio.name}", t, text])
-    # save_csv("csv/whisper.csv", rows)
+        d, nd, diff = get_text_distance(original[audio.stem], text)
+        rows.append([audio.name, round(t, 3), text, d, round(nd,3), diff])
+    save_csv("csv/whisper.csv", rows)
 
+
+def run_test_audios():
+    model = load_model()
+    lang = "zh"
+    audios = Path("/Users/jeqin/work/code/TestTranslator/tests/test_data/test_audios/")
+    rows = [["file_name", "time", "inference_result"]]
+    for audio in sorted(audios.glob(f"*{lang}*/*.wav")):
+        print(audio)
+        t1 = time.time()
+        output = model.transcribe(str(audio), language=lang, initial_prompt="以下是普通话句子，这是一段会议内容。")# initial_prompt="这是一段中文的会议内容。"
+        t = time.time() - t1
+        print("inference time:", t)
+        text = " ".join([a.text for a in output])
+        print(text)
+        rows.append([f"{audio.parent.name}/{audio.name}", round(t, 3), text])
+    save_csv("csv/whisper.csv", rows)
 if __name__ == '__main__':
-    main()
+    run_recordings()

temp.py

@@ -1,4 +1,14 @@
-text ="""
-{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0].role == 'system' %}\n        {{- messages[0].content + '\\n\\n' }}\n    {%- endif %}\n    {{- \"# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0].role == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0].content + '<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\n{%- for message in messages[::-1] %}\n    {%- set index = (messages|length - 1) - loop.index0 %}\n    {%- if ns.multi_step_tool and message.role == \"user\" and message.content is string and not(message.content.startswith('<tool_response>') and message.content.endswith('</tool_response>')) %}\n        {%- set ns.multi_step_tool = false %}\n        {%- set ns.last_query_index = index %}\n    {%- endif %}\n{%- endfor %}\n{%- for message in messages %}\n    {%- if message.content is string %}\n        {%- set content = message.content %}\n    {%- else %}\n        {%- set content = '' %}\n    {%- endif %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) %}\n        {{- '<|im_start|>' + message.role + '\\n' + content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {%- set reasoning_content = '' %}\n        {%- if message.reasoning_content is string %}\n            {%- set reasoning_content = message.reasoning_content %}\n        {%- else %}\n            {%- if '</think>' in content %}\n                {%- set reasoning_content = content.split('</think>')[0].rstrip('\\n').split('<think>')[-1].lstrip('\\n') %}\n                {%- set content = content.split('</think>')[-1].lstrip('\\n') %}\n            {%- endif %}\n        {%- endif %}\n        {%- if loop.index0 > ns.last_query_index %}\n            {%- if loop.last or (not loop.last and reasoning_content) %}\n                {{- '<|im_start|>' + message.role + '\\n<think>\\n' + reasoning_content.strip('\\n') + '\\n</think>\\n\\n' + content.lstrip('\\n') }}\n            {%- else %}\n                {{- '<|im_start|>' + message.role + '\\n' + content }}\n            {%- endif %}\n        {%- else %}\n            {{- '<|im_start|>' + message.role + '\\n' + content }}\n        {%- endif %}\n        {%- if message.tool_calls %}\n            {%- for tool_call in message.tool_calls %}\n                {%- if (loop.first and content) or (not loop.first) %}\n                    {{- '\\n' }}\n                {%- endif %}\n                {%- if tool_call.function %}\n                    {%- set tool_call = tool_call.function %}\n                {%- endif %}\n                {{- '<tool_call>\\n{\"name\": \"' }}\n                {{- tool_call.name }}\n                {{- '\", \"arguments\": ' }}\n                {%- if tool_call.arguments is string %}\n                    {{- tool_call.arguments }}\n                {%- else %}\n                    {{- tool_call.arguments | tojson }}\n                {%- endif %}\n                {{- '}\\n</tool_call>' }}\n            {%- endfor %}\n        {%- endif %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if loop.first or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}
-"""
-print(text)
+TEXT_FILE = '/Users/jeqin/work/code/TestTranslator/tests/test_data/text/test_asr_zh.txt'
+index = 1
+with open(TEXT_FILE, encoding='utf-8') as f:
+    for line in f:
+        line = line.strip()
+        # print(line)
+        if not line:
+            continue
+        if line.startswith('#'):
+            # print(line)
+            continue
+        line = f"{index}. {line}"
+        print(line)
+        index += 1

tests/test_data/recordings/1.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77f6f1006e789f69f30f9047d5d93f2cbc58012f1a41a21ebfc12b93c2de7d89
+size 141854

tests/test_data/recordings/10.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ce3be6f2ae27a19ee56eafb95ac3ca62b7946c4733e8aa542e74f36d06b036b
+size 184664

tests/test_data/recordings/11.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6720957bd8bcba4515cc3e416d9898c38f2dbdb377697ef69a528e621e11d0a1
+size 158234

tests/test_data/recordings/12.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c686cd5322384e36747647f228cd023a77bc323cc26fab9f7fcd351f93019dc3
+size 201614

tests/test_data/recordings/13.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:baf0c19d4729ecba8057a2292cdb117ec9c4836aaa5aabbb9def63ef1d7ffbb3
+size 201674

tests/test_data/recordings/14.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44baf4bdd176224377b4318400d5ac8a4947517fe09b0e16619d63d77b18a631
+size 321194

tests/test_data/recordings/15.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5abf82f0fa3ce6737f7cdae83931b78295bf2adc48f394d1fa6c7193c76e879
+size 252284

tests/test_data/recordings/16.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3323eeefe1f1219cf2db94a41e0370f699f66e495c390b5d0a8576748955a220
+size 278594

tests/test_data/recordings/17.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8654234ed4a01ef2b0b32779757da30c5c6929325977c0d7640d40216d12abd5
+size 381464

tests/test_data/recordings/18.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ece527a8f191c88bda2800715898ae5e3c2ec13b86d43467ec9e37a53b8e77c
+size 284024

tests/test_data/recordings/19.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5d57a7b312033cc19cc7f7a6db06b01b6d37e191a58588b9b32a7ab98032aa3
+size 367964

tests/test_data/recordings/2.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31be8ce54c599118787494d88e481e511dab05d45d1a7383710ceecf3db7569d
+size 149924

tests/test_data/recordings/20.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76acbe509263e26fd076bab325cde6b45c52d792a562180ddb9b5a4da62beec7
+size 274964

tests/test_data/recordings/21.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e143f1d398c8f144324767474805d69b9a1f285734f354d1ff10824f4f869b70
+size 313754

tests/test_data/recordings/22.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf10a0bd8f9b202a5fc32f467d7a55c4029675a28f428970a2149f50a315f112
+size 272714

tests/test_data/recordings/23.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c17c5316b39779bcb84d0aa1593c05fc22907697f6d6c7236c7a7260166e406b
+size 266204

tests/test_data/recordings/24.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb1dba9272cd68902b20e705ce687b6cce5e6ef79a5427b7a1bcbb0532137227
+size 297314

tests/test_data/recordings/25.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b89d3440b113f0db28ec55776d3ca97cdc430ea9d1a13023bfec0d5aba5cabe9
+size 270434

tests/test_data/recordings/26.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a08d7996a8485369483e7e6f2e19e7c21dbdec91df3bae74a4cba360eb80d9e4
+size 150704

tests/test_data/recordings/27.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae80a1a961185e18c9b93216b3ce1a5baee9e3aee0fbcf4114a0319c9fce556a
+size 174524

tests/test_data/recordings/28.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9293d2d95d78ce9392fd49e220798eb2ef79a7b361e5c9aaa00451125d79c721
+size 183014

tests/test_data/recordings/29.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a9f941054691b17c3d6320b1bd4ae0468ab4f335d667d87ae25ef2b4ba251f1
+size 196994

tests/test_data/recordings/3.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78c500e9787d1882bde1e3d7fe5ab48bffebc69a503e99f0f6ba7d011941e687
+size 242084

tests/test_data/recordings/30.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32fbcac1e31b4fbec5a758a40ac01c477b3d79b79c78195134d17f51ebaf5023
+size 162194

tests/test_data/recordings/31.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4bb89d406d30a1cc77aea7f7d2162fc9db33b33c6f4d6258b09c0018c2b3e37
+size 190634

tests/test_data/recordings/32.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3632d894647cdae67fecb4703fcf7ce1030cd905580e20fbeca5cc51d2b0e40c
+size 268694

tests/test_data/recordings/33.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e52805deabef8843e1f75ac8ae651f766dd5e0225981a8b843109e73b0f522db
+size 247094

tests/test_data/recordings/34.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9008cf27e698797c14dc9f5594952a8a84007c8a6f2f674dbf69bed276d392a
+size 248234

tests/test_data/recordings/35.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e15a82a43aa78ab86983c85883ed49ffbb99fc87cccf33022eb6bcb8d9eea88c
+size 257864

tests/test_data/recordings/36.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:020ed5cb9eb3a00e4fdb9408783d36cc758534e0c3263a0896995faebdcf5535
+size 274394

tests/test_data/recordings/37.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f4d2d0f66885e6bee03bf4806cd2cf760ebec7b120bdbcb49ac46c849e0c1a4
+size 295964

tests/test_data/recordings/38.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a54649f65aabfdd20fe2df2c4b1f6e294e773540cb63b65503c0d1732c990319
+size 277334

tests/test_data/recordings/39.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae68f060247ea11ee422bffa01d1e444fa323b7e2105a63079ca49cf2d9e0df7
+size 347624

tests/test_data/recordings/4.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:113d7173b16bd2d921d0e63b8a1f39fae1450da68b4b6e835d8f6bea405bf25e
+size 227264

tests/test_data/recordings/40.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb33cf5324157fcc1a81398f467c75ca0c31a697b0eb378cfef37c237af3e875
+size 277994

tests/test_data/recordings/41.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc3ef23125d4c5246ad4bd3e41b4f842356ecfd186ff6112b1f12a2474e912ec
+size 258284

tests/test_data/recordings/42.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9efd63be189ba2eb143ad8cabb03c1088f113e3b25bd586f5998669bca2acb10
+size 159404

tests/test_data/recordings/43.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:343c18a93f85637f7872953f98d1fc5a03d37c2271b147b53a4f68f0126d2fad
+size 556814

tests/test_data/recordings/44.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f52275b9d21e17e527a9dff2e7f102210764d1f305e269a1f336c2ddb80fcf9
+size 246134

tests/test_data/recordings/45.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fe73673d6d8ed05031ffdaac3015419ee3a495022d515e0161867fe49da139b
+size 296894

tests/test_data/recordings/46.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:716f798077ff25b2f8bc7620ba70b45f6a51152b70079688eaa6dae18b90f464
+size 279404

tests/test_data/recordings/47.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66866edef043eedf895fdeac5ea025291be2e68e64819355b56c6588a51bb136
+size 285074