Add new input m3u8 URL | E.g.: from kvf.fo or logting.fo (#1)

- Add new input: m3u8 URL | E.g.: from kvf.fo or logting.fo (a2270aa2db9b23edc1160c9ebff5a096ff5a770a)

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 import time
 from transformers import pipeline
 import torch
+import ffmpeg  # Make sure it's ffmpeg-python
 
 # Check if GPU is available
 use_gpu = torch.cuda.is_available()
@@ -14,13 +15,27 @@ if use_gpu:
 else:
     p = pipeline("automatic-speech-recognition", 
              model="carlosdanielhernandezmena/wav2vec2-large-xlsr-53-faroese-100h")
-    
 
-def transcribe(audio, state="", uploaded_audio=None):
+
+def extract_audio_from_m3u8(url):
+    try:
+        output_file = "output_audio.aac"
+        ffmpeg.input(url).output(output_file).run(overwrite_output=True)
+        return output_file
+    except Exception as e:
+        return f"An error occurred: {e}"
+
+
+def transcribe(audio, state="", uploaded_audio=None, m3u8_url=""):
+    if m3u8_url:
+        audio = extract_audio_from_m3u8(m3u8_url)
+
     if uploaded_audio is not None:
         audio = uploaded_audio
+
     if not audio:
         return state, state  # Return a meaningful message
+
     try:
         time.sleep(3)
         text = p(audio, chunk_length_s= 50)["text"]
@@ -40,10 +55,11 @@ demo = gr.Interface(
     inputs=[
         gr.components.Audio(source="microphone", type="filepath"),
         'state',
-        gr.components.Audio(label="Upload Audio File", type="filepath", source="upload")
+        gr.components.Audio(label="Upload Audio File", type="filepath", source="upload"),
+        gr.components.Textbox(label="m3u8 URL | E.g.: from kvf.fo or logting.fo")
     ],
     outputs=[
-        "textbox",
+        gr.components.Textbox(type="text"),
         "state"
     ],
 

requirements.txt

@@ -1,2 +1,5 @@
 torch 
 transformers
+gradio
+ffmpeg-python
+ffmpeg