Simplest speech to text example

.gitignore

@@ -0,0 +1 @@
+/venv

app.py

@@ -0,0 +1,28 @@
+from transformers import pipeline, AutoFeatureExtractor, AutoTokenizer, Wav2Vec2ForCTC
+import gradio as gr
+import time
+
+model_id = 'comodoro/wav2vec2-xls-r-300m-cs-250'
+feature_extractor = AutoFeatureExtractor.from_pretrained(model_id)
+model = Wav2Vec2ForCTC.from_pretrained(model_id)
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+
+p = pipeline("automatic-speech-recognition", chunk_length_s=5, model=model, 
+    tokenizer=tokenizer, feature_extractor=feature_extractor)
+
+def transcribe(audio, state=""):
+    time.sleep(2)
+    text = p(audio)["text"]
+    state += text + " "
+    return state
+
+with gr.Blocks() as blocks:
+  audio = gr.Audio(source="microphone", type="filepath",
+      label='Pokud je to třeba, povolte mikrofon pro tuto stránku, \
+ klikněte na Record from microphone, po dokončení nahrávání na Stop recording a poté na Rozpoznat')
+  btn = gr.Button('Rozpoznat')
+  output = gr.Textbox(show_label=False)
+  btn.click(fn=transcribe, inputs=[audio,], 
+      outputs=[output,])
+
+blocks.launch(enable_queue=True, debug=True) 

packages.txt

@@ -0,0 +1 @@
+ffmpeg

requirements.txt

@@ -0,0 +1,2 @@
+torch
+transformers