first try

.gitignore

@@ -0,0 +1 @@
+.idea/

app.py

@@ -0,0 +1,51 @@
+from transformers import pipeline
+from pytube import YouTube
+
+import gradio as gr
+import librosa
+
+import hopsworks
+
+project = hopsworks.login()
+fs = project.get_feature_store()
+
+dataset_api = project.get_dataset_api()
+
+dataset_api.download("Resources/titanic/images/latest_titanic.png", overwrite=True) # change link 
+# dataset_api.download("Resources/images/deadImage.png", overwrite=True) # change link 
+
+
+# pipe = pipeline(model="fimster/whisper-small-sv-SE") # change model
+# pipe = pipeline(model="ayberkuckun/whisper-small-sv-SE")
+pipe = pipeline(model="openai/whisper-small")
+
+
+def transcribe(url):
+    selected_video = YouTube(url)
+
+    try:
+        audio = selected_video.streams.filter(only_audio=True, file_extension='mp4')[0]
+    except:
+        raise Exception("Can't find an mp4 audio.")
+
+    audio.download(filename="audio.mp4")
+
+    speech_array, sr = librosa.load("audio.mp4", sr=16000)
+
+    output = pipe(speech_array[:sr*30])
+
+    return output["text"]
+
+
+iface = gr.Interface(
+    fn=transcribe, 
+    inputs=gr.Textbox("https://www.youtube.com/watch?v=n9g12Xm9UJM", label="Paste a YouTube video URL"),
+    outputs=[gr.Textbox(label="Only the first approximately 30 sec will be transcripted"),
+            gr.Image("latest_titanic.png", label="Model Scores")],
+              # gr.Image("deadImage.png", elem_id="predicted-img", label="Model Scores")],
+    title="Whisper Small Swedish",
+    description="Realtime demo for Swedish speech recognition using a fine-tuned Whisper small model.",
+    allow_flagging="never"
+)
+
+iface.launch()

requirements.txt

@@ -0,0 +1,5 @@
+hopsworks
+librosa
+gradio
+git+https://github.com/huggingface/transformers
+git+https://github.com/pytube/pytube