B64

.history/app_20230718133117.py

@@ -0,0 +1,24 @@
+import gradio as gr
+from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
+import torch
+import phonemizer
+import librosa
+
+processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+
+waveform, sample_rate = librosa.load('harvard.wav', sr=16000) # Downsample 44.1kHz to 8kHz
+
+input_values = processor(waveform, sampling_rate=sample_rate, return_tensors="pt").input_values
+
+with torch.no_grad():
+   logits = model(input_values).logits
+
+predicted_ids = torch.argmax(logits, dim=-1)
+transcription = processor.batch_decode(predicted_ids)
+
+def showTranscription(transcription):
+   return transcription
+
+iface = gr.Interface(fn=showTranscription, inputs="text", outputs="text")
+iface.launch()

.history/app_20230718133128.py

@@ -0,0 +1,24 @@
+import gradio as gr
+from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
+import torch
+import phonemizer
+import librosa
+
+processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+
+waveform, sample_rate = librosa.load('harvard.wav', sr=16000) # Downsample 44.1kHz to 8kHz
+
+input_values = processor(waveform, sampling_rate=sample_rate, return_tensors="pt").input_values
+
+with torch.no_grad():
+   logits = model(input_values).logits
+
+predicted_ids = torch.argmax(logits, dim=-1)
+transcription = processor.batch_decode(predicted_ids)
+
+def showTranscription(transcription):
+   return transcription
+
+iface = gr.Interface(fn=showTranscription, inputs="text", outputs="text")
+iface.launch()

.history/app_20230718133340.py

@@ -0,0 +1,32 @@
+import gradio as gr
+from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
+import torch
+import phonemizer
+import librosa
+import base64
+
+
+processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+
+waveform, sample_rate = librosa.load(
+    "harvard.wav", sr=16000
+)  # Downsample 44.1kHz to 8kHz
+
+input_values = processor(
+    waveform, sampling_rate=sample_rate, return_tensors="pt"
+).input_values
+
+with torch.no_grad():
+    logits = model(input_values).logits
+
+predicted_ids = torch.argmax(logits, dim=-1)
+transcription = processor.batch_decode(predicted_ids)
+
+
+def showTranscription(transcription):
+    return transcription
+
+
+iface = gr.Interface(fn=showTranscription, inputs="text", outputs="text")
+iface.launch()

.history/app_20230718133558.py

@@ -0,0 +1,33 @@
+import gradio as gr
+from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
+import torch
+import phonemizer
+import librosa
+import base64
+
+
+def lark(audioAsB64):
+    with open("audio.wav", "wb") as preWaveform:
+        preWaveform.write(base64.b64encode())
+
+
+processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+
+waveform, sample_rate = librosa.load(
+    "harvard.wav", sr=16000
+)  # Downsample 44.1kHz to 8kHz
+
+input_values = processor(
+    waveform, sampling_rate=sample_rate, return_tensors="pt"
+).input_values
+
+with torch.no_grad():
+    logits = model(input_values).logits
+
+predicted_ids = torch.argmax(logits, dim=-1)
+transcription = processor.batch_decode(predicted_ids)
+
+
+iface = gr.Interface(fn=lark, inputs="text", outputs="text")
+iface.launch()

.history/app_20230718133701.py

@@ -0,0 +1,36 @@
+import gradio as gr
+from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
+import torch
+import phonemizer
+import librosa
+import base64
+
+
+def lark(audioAsB64):
+    # convert b64 audio to wav
+    with open("audio.wav", "wb") as preWaveform:
+        preWaveform.write(base64.b64encode())
+
+    # processing
+    processor = Wav2Vec2Processor.from_pretrained(
+        "facebook/wav2vec2-xlsr-53-espeak-cv-ft"
+    )
+    model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+
+    waveform, sample_rate = librosa.load(
+        "harvard.wav", sr=16000
+    )  # Downsample 44.1kHz to 8kHz
+
+    input_values = processor(
+        waveform, sampling_rate=sample_rate, return_tensors="pt"
+    ).input_values
+
+    with torch.no_grad():
+        logits = model(input_values).logits
+
+    predicted_ids = torch.argmax(logits, dim=-1)
+    transcription = processor.batch_decode(predicted_ids)
+
+
+iface = gr.Interface(fn=lark, inputs="text", outputs="text")
+iface.launch()

.history/app_20230718133728.py

@@ -0,0 +1,38 @@
+import gradio as gr
+from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
+import torch
+import phonemizer
+import librosa
+import base64
+
+
+def lark(audioAsB64):
+    # convert b64 audio to wav
+    with open("audio.wav", "wb") as preWaveform:
+        preWaveform.write(base64.b64encode())
+
+    # processing
+    processor = Wav2Vec2Processor.from_pretrained(
+        "facebook/wav2vec2-xlsr-53-espeak-cv-ft"
+    )
+    model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+
+    waveform, sample_rate = librosa.load(
+        "harvard.wav", sr=16000
+    )  # Downsample 44.1kHz to 8kHz
+
+    input_values = processor(
+        waveform, sampling_rate=sample_rate, return_tensors="pt"
+    ).input_values
+
+    with torch.no_grad():
+        logits = model(input_values).logits
+
+    predicted_ids = torch.argmax(logits, dim=-1)
+    transcription = processor.batch_decode(predicted_ids)
+
+    return transcription
+
+
+iface = gr.Interface(fn=lark, inputs="text", outputs="text")
+iface.launch()

.history/app_20230718134339.py

@@ -0,0 +1,38 @@
+import gradio as gr
+from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
+import torch
+import phonemizer
+import librosa
+import base64
+
+
+def lark(audioAsB64):
+    # convert b64 audio to wav
+    with open("audio.wav", "wb") as preWaveform:
+        preWaveform.write(base64.b64encode())
+
+    # processing
+    processor = Wav2Vec2Processor.from_pretrained(
+        "facebook/wav2vec2-xlsr-53-espeak-cv-ft"
+    )
+    model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-xlsr-53-espeak-cv-ft")
+
+    waveform, sample_rate = librosa.load(
+        "harvard.wav", sr=16000
+    )  # Downsample 44.1kHz to 8kHz
+
+    input_values = processor(
+        waveform, sampling_rate=sample_rate, return_tensors="pt"
+    ).input_values
+
+    with torch.no_grad():
+        logits = model(input_values).logits
+
+    predicted_ids = torch.argmax(logits, dim=-1)
+    transcription = processor.batch_decode(predicted_ids)
+
+    return transcription
+
+
+iface = gr.Interface(fn=lark, inputs="text", outputs="text")
+iface.launch()

.history/packages_20230718132842.txt

@@ -0,0 +1 @@
+espeak

.history/requirements_20230718132835.txt

@@ -0,0 +1,4 @@
+phonemizer
+librosa
+transformers
+torch

.history/requirements_20230718133331.txt

@@ -0,0 +1,5 @@
+phonemizer
+librosa
+transformers
+torch
+base64

.history/requirements_20230718134813.txt

@@ -0,0 +1,4 @@
+phonemizer
+librosa
+transformers
+torch

.history/requirements_20230718134828.txt

@@ -0,0 +1,4 @@
+phonemizer
+librosa
+transformers
+torch

.vscode/settings.json

@@ -0,0 +1,6 @@
+{
+  "[python]": {
+    "editor.defaultFormatter": "ms-python.black-formatter"
+  },
+  "python.formatting.provider": "none"
+}

requirements.txt

@@ -1,5 +1,4 @@
 phonemizer
 librosa
 transformers
-torch
-base64
+torch