Update app.py

app.py

@@ -2,149 +2,85 @@ import os
 import tempfile
 from pathlib import Path
 import gradio as gr
+from faster_whisper import WhisperModel
 import ffmpeg
-import whisper
+
+# -------- Configuration --------
+MODEL_NAME = "small"  # choices: tiny, base, small, medium, large-v3
+DEVICE = "cuda" if os.environ.get("USE_CUDA", "0") == "1" else "cpu"
+
+# Load model once
+print(f"🚀 Loading Faster-Whisper model: {MODEL_NAME} on {DEVICE}")
+model = WhisperModel(MODEL_NAME, device=DEVICE, compute_type="float16" if DEVICE == "cuda" else "int8")
 
 # -------- Helper functions --------
 def _format_timestamp(seconds: float) -> str:
+    """Return hh:mm:ss,ms timestamp format."""
+    if seconds is None:
+        return "00:00:00,000"
     ms = int(round(seconds * 1000))
-    hours = ms // 3600000
-    ms %= 3600000
-    minutes = ms // 60000
-    ms %= 60000
-    seconds = ms // 1000
-    millis = ms % 1000
-    return f"{hours:02d}:{minutes:02d}:{seconds:02d},{millis:03d}"
-
-
-def segments_to_srt(segments: list) -> str:
-    lines = []
-    for i, seg in enumerate(segments, start=1):
-        start_ts = _format_timestamp(seg["start"])
-        end_ts = _format_timestamp(seg["end"])
-        text = seg["text"].strip().replace("\n", " ")
-        if text:
-            lines.append(f"{i}\n{start_ts} --> {end_ts}\n{text}\n")
-    return "\n".join(lines)
-
-
-# -------- Config --------
-MODEL_NAME = "base"
-OUTPUT_DIR = Path("outputs/subtitles")
-OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
-
-print(f"Loading Whisper model '{MODEL_NAME}'...")
-model = whisper.load_model(MODEL_NAME)
-print("Model loaded successfully!")
-
-
-# -------- Core functions --------
-def extract_audio(input_path: str, out_path: str):
-    """Extract mono 16 kHz WAV using ffmpeg"""
-    (
-        ffmpeg
-        .input(input_path)
-        .output(out_path, format="wav", acodec="pcm_s16le", ac=1, ar="16000")
-        .overwrite_output()
-        .run(quiet=True)
-    )
-
-
-def transcribe_file_to_srt(file_obj, language="en"):
-    tmp_dir = Path(tempfile.mkdtemp(prefix="subgen_"))
-    input_path = tmp_dir / Path(file_obj.name).name
-    with open(input_path, "wb") as f:
-        f.write(file_obj.read())
-
-    audio_path = tmp_dir / "audio.wav"
-    extract_audio(str(input_path), str(audio_path))
-
-    result = model.transcribe(str(audio_path), language=language)
-    segments = []
-    for i, seg in enumerate(result["segments"]):
-        segments.append({
-            "start": seg["start"],
-            "end": seg["end"],
-            "text": seg["text"]
-        })
-
-    srt_text = segments_to_srt(segments)
-    output_path = OUTPUT_DIR / f"{Path(file_obj.name).stem}.srt"
-    output_path.write_text(srt_text, encoding="utf-8")
-
-    return str(output_path), "✅ Subtitles generated successfully!"
-
-
-# -------- UI Styling --------
-def build_style(theme="light"):
-    if theme == "dark":
-        bg = "#0f2027"
-        color = "#ffffff"
-        button = "#00adb5"
-    else:
-        bg = "#f0f2f5"
-        color = "#000000"
-        button = "#0077ff"
-
-    return f"""
-    <style>
-    body {{
-        background: {bg};
-        color: {color};
-        font-family: 'Poppins', sans-serif;
-        transition: background 0.5s, color 0.5s;
-    }}
-    .gr-button {{
-        background-color: {button} !important;
-        color: white !important;
-        font-weight: bold;
-        border-radius: 10px !important;
-    }}
-    .gr-button:hover {{
-        filter: brightness(1.2);
-    }}
-    </style>
-    """
+    hours = ms // 3_600_000
+    minutes = (ms % 3_600_000) // 60_000
+    seconds = (ms % 60_000) // 1000
+    milliseconds = ms % 1000
+    return f"{hours:02d}:{minutes:02d}:{seconds:02d},{milliseconds:03d}"
+
+
+def transcribe(audio_file):
+    """Transcribe uploaded audio file and return text + SRT."""
+    try:
+        # Convert to wav if needed (ensures consistency)
+        with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tmp_wav:
+            (
+                ffmpeg
+                .input(audio_file)
+                .output(tmp_wav.name, format="wav", acodec="pcm_s16le", ac=1, ar="16k")
+                .overwrite_output()
+                .run(quiet=True)
+            )
+            wav_path = tmp_wav.name
+
+        # Run transcription
+        segments, info = model.transcribe(wav_path, beam_size=5)
+        text_output = ""
+        srt_output = ""
+
+        for i, segment in enumerate(segments, start=1):
+            start = _format_timestamp(segment.start)
+            end = _format_timestamp(segment.end)
+            srt_output += f"{i}\n{start} --> {end}\n{segment.text.strip()}\n\n"
+            text_output += segment.text.strip() + " "
+
+        return text_output.strip(), srt_output
+
+    except Exception as e:
+        return f"Error: {str(e)}", ""
 
 
 # -------- Gradio UI --------
-with gr.Blocks(title="AI Subtitle Generator") as demo:
-    theme_state = gr.State("light")
-
-    style_html = gr.HTML(build_style("light"))
-    gr.HTML("<h1 style='text-align:center;'>🎬 AI Subtitle Generator</h1>")
-    gr.HTML("<p style='text-align:center;'>Upload a video or audio file to generate English <b>.srt</b> subtitles.</p>")
-
-    with gr.Row():
-        input_file = gr.File(label="Upload video/audio file")
-        output_file = gr.File(label="Download .srt file")
+def build_ui():
+    with gr.Blocks(title="🎙️ Faster-Whisper Transcriber") as app:
+        gr.Markdown("# 🎧 Fast & Accurate Speech-to-Text using Faster-Whisper")
+        gr.Markdown("Upload or record an audio file to get instant transcription.")
 
-    status_box = gr.Textbox(label="Status", interactive=False)
+        with gr.Row():
+            audio_input = gr.Audio(sources=["upload", "microphone"], type="filepath", label="Audio Input")
 
-    with gr.Row():
-        generate_btn = gr.Button("🚀 Generate Subtitles")
-        clear_btn = gr.Button("🧹 Clear Chat")
-        theme_btn = gr.Button("🌗 Toggle Theme")
+        with gr.Row():
+            text_output = gr.Textbox(label="Transcribed Text", lines=6)
+            srt_output = gr.Textbox(label="SRT Subtitle", lines=6)
 
-    # Button logic
-    def on_generate(file):
-        if not file:
-            return None, "⚠️ Please upload a file first!"
-        srt_path, msg = transcribe_file_to_srt(file)
-        return srt_path, msg
+        transcribe_btn = gr.Button("Transcribe")
 
-    def on_clear():
-        return None, None, ""
+        transcribe_btn.click(fn=transcribe, inputs=audio_input, outputs=[text_output, srt_output])
 
-    def on_theme(current):
-        new_theme = "dark" if current == "light" else "light"
-        return new_theme, gr.update(value=build_style(new_theme))
+        gr.Markdown("---")
+        gr.Markdown("Powered by **Faster-Whisper** ⚡")
 
-    generate_btn.click(on_generate, inputs=[input_file], outputs=[output_file, status_box])
-    clear_btn.click(on_clear, outputs=[input_file, output_file, status_box])
-    theme_btn.click(on_theme, inputs=[theme_state], outputs=[theme_state, style_html])
+    return app
 
-    gr.HTML("<p style='text-align:center;font-size:14px;opacity:0.6;'>✨ Built with OpenAI Whisper + Gradio</p>")
 
+# -------- Launch --------
 if __name__ == "__main__":
-    demo.launch()
+    app = build_ui()
+    app.launch(server_name="0.0.0.0", server_port=7860)