Update app.py

app.py

@@ -1,9 +1,9 @@
 # -*- coding: utf-8 -*-
 """ 
-ROBOTSMALI — Sous-titrage Bambara (V5.3 - Production)
+ROBOTSMALI — Sous-titrage Bambara (V5.5 - Production Ready)
 - Vidéo d'exemple : examples/MARALINKE.mp4
 - Correction AttributeError: Gradio Div -> Column/HTML
-- Correction Codec Webcam : VP8 -> H.264
+- Correction Codec Webcam : VP8 -> H.264 (Stabilisation forcée)
 """
 import os
 import shlex
@@ -22,12 +22,27 @@ from huggingface_hub import snapshot_download
 from nemo.collections import asr as nemo_asr
 import gradio as gr
 
+# ---------------------------- # VÉRIFICATION DIAGNOSTIC # ----------------------------
+print("--- DIAGNOSTIC DES FICHIERS ---")
+example_path = "examples/MARALINKE.mp4"
+if os.path.exists(example_path):
+    print(f"✅ SUCCÈS : {example_path} est bien présent.")
+else:
+    print(f"❌ ERREUR : {example_path} est introuvable !")
+    if os.path.exists("examples"):
+        print(f"Contenu réel du dossier examples/ : {os.listdir('examples')}")
+    else:
+        print("Le dossier 'examples' n'existe pas à la racine du projet.")
+print("-------------------------------")
+
 # ---------------------------- # CONFIGURATION # ----------------------------
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 random.seed(1234)
 np.random.seed(1234)
 torch.manual_seed(1234)
 
+SEGMENT_DURATION = 10.0
+
 MODELS = {
     "Soloni V1 (RNNT)":        ("RobotsMali/soloni-114m-tdt-ctc-v1", "rnnt"),
     "Soloni V0 (RNNT)":        ("RobotsMali/soloni-114m-tdt-ctc-v0", "rnnt"),
@@ -37,10 +52,7 @@ MODELS = {
     "QuartzNet V0 (CTC-char)": ("RobotsMali/stt-bm-quartznet15x5-v0", "ctc_char"),
 }
 
-# Mise à jour avec le nom simplifié
-VIDEO_EXAMPLES = [
-    ["examples/MARALINKE.mp4", "Soloba V1 (CTC)"]
-]
+VIDEO_EXAMPLES = [[example_path, "Soloba V1 (CTC)"]]
 
 _cache = {}
 
@@ -63,6 +75,7 @@ def load_model(name):
     repo, mode = MODELS[name]
     folder = snapshot_download(repo, local_dir_use_symlinks=False)
     nemo_file = next((os.path.join(folder, f) for f in os.listdir(folder) if f.endswith(".nemo")), None)
+    
     if mode == "rnnt":
         model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.restore_from(nemo_file)
     elif mode == "ctc_char":
@@ -70,14 +83,16 @@ def load_model(name):
     else:
         try: model = nemo_asr.models.EncDecCTCModelBPE.restore_from(nemo_file)
         except: model = nemo_asr.models.EncDecCTCModel.restore_from(nemo_file)
+
     model.to(DEVICE).eval()
     _cache[name] = model
     return model
 
 def extract_audio(video_path, out_wav):
+    """Stabilisation pour flux webcam et extraction audio."""
     tmp_fd, stabilized_mp4 = tempfile.mkstemp(suffix="_stabilized.mp4")
     os.close(tmp_fd)
-    # Réencodage H.264 pour supporter le flux webcam
+    # Correction WebM/Webcam : réencodage libx264 forcé
     run_cmd(f'ffmpeg -hide_banner -loglevel error -y -i {shlex.quote(video_path)} -c:v libx264 -preset ultrafast -crf 23 -c:a aac {shlex.quote(stabilized_mp4)}')
     run_cmd(f'ffmpeg -hide_banner -loglevel error -y -i {shlex.quote(stabilized_mp4)} -vn -ac 1 -ar 16000 -f wav {shlex.quote(out_wav)}')
     if os.path.exists(stabilized_mp4): os.remove(stabilized_mp4)
@@ -92,14 +107,14 @@ def clean_audio(wav_path):
     sf.write(clean_path, audio, 16000)
     return clean_path, audio, 16000
 
-# ---------------------------- # PIPELINE # ----------------------------
+# ---------------------------- # PIPELINE PRINCIPAL # ----------------------------
 
 def pipeline(video_input, model_name):
     try:
-        if not video_input: return "❌ Vidéo introuvable", None
+        if not video_input: return "❌ Vidéo introuvable. Veuillez réessayer.", None
         video_path = video_input
         
-        yield "⏳ Phase 1 : Extraction audio...", None
+        yield "⏳ Phase 1/3 : Analyse du fichier et extraction audio...", None
         with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as tf:
             wav_path = tf.name
         
@@ -107,15 +122,16 @@ def pipeline(video_input, model_name):
         clean_wav, audio, sr = clean_audio(wav_path)
         duration = ffprobe_duration(video_path) or (len(audio)/sr)
         
-        yield f"⏳ Phase 2 : Transcription IA ({model_name})...", None
+        yield f"⏳ Phase 2/3 : Transcription IA avec {model_name}...", None
         model = load_model(model_name)
-        text = model.transcribe([clean_wav])[0]
-        text_str = text.text if hasattr(text, 'text') else str(text)
-        words = [w for w in text_str.split() if len(w) > 1]
+        text_out = model.transcribe([clean_wav])[0]
+        text_str = text_out.text if hasattr(text_out, 'text') else str(text_out)
         
-        if not words: return "⚠️ Pas de parole détectée", None
+        words = [w for w in text_str.split() if len(w) > 1]
+        if not words: return "⚠️ Aucune parole détectée dans la vidéo.", None
         
-        yield "⏳ Phase 3 : Incrustation des sous-titres...", None
+        yield "⏳ Phase 3/3 : Incrustation des sous-titres...", None
+        # Heuristique d'alignement simple
         subs = []
         chunk_size = 7
         for i in range(0, len(words), chunk_size):
@@ -125,13 +141,13 @@ def pipeline(video_input, model_name):
             subs.append((s, e, "\n".join(textwrap.wrap(" ".join(chunk), 40))))
             
         res_v = burn(video_path, subs)
-        yield "✅ Succès !", res_v
+        yield "✅ Succès ! Votre vidéo est prête.", res_v
     except Exception as e:
         traceback.print_exc()
-        yield f"❌ Erreur: {str(e)}", None
+        yield f"❌ Erreur : {str(e)}", None
 
 def burn(video_path, subs):
-    out_path = "RobotsMali_Final.mp4"
+    out_path = "RobotsMali_Subtitled.mp4"
     with tempfile.NamedTemporaryFile(suffix=".srt", mode="w", encoding="utf-8", delete=False) as tf:
         for idx, (start, end, text) in enumerate(subs, 1):
             def t_srt(sec):
@@ -139,8 +155,9 @@ def burn(video_path, subs):
                 return f"{h:02}:{m:02}:{s:02},{ms:03}"
             tf.write(f"{idx}\n{t_srt(start)} --> {t_srt(end)}\n{text}\n\n")
         srt_name = tf.name
-    vf = f"subtitles={shlex.quote(srt_name)}:force_style='Fontsize=22,PrimaryColour=&HFFFFFF&,OutlineColour=&H000000&'"
-    run_cmd(f'ffmpeg -hide_banner -loglevel error -y -i {shlex.quote(video_path)} -vf {shlex.quote(vf)} -c:v libx264 -crf 23 -c:a aac {shlex.quote(out_path)}')
+    
+    vf = f"subtitles={shlex.quote(srt_name)}:force_style='Fontsize=24,PrimaryColour=&HFFFFFF&,OutlineColour=&H000000&'"
+    run_cmd(f'ffmpeg -hide_banner -loglevel error -y -i {shlex.quote(video_path)} -vf {shlex.quote(vf)} -c:v libx264 -preset fast -crf 23 -c:a aac {shlex.quote(out_path)}')
     os.remove(srt_name)
     return out_path
 
@@ -148,45 +165,42 @@ def burn(video_path, subs):
 
 custom_css = """
 body { background-color: #0b0e14; }
-.gradio-container { background: rgba(17, 25, 40, 0.8) !important; backdrop-filter: blur(12px); border-radius: 20px; border: 1px solid rgba(255, 255, 255, 0.1); }
-#title-container { text-align: center; padding: 20px; }
-.gr-button-primary { background: linear-gradient(135deg, #059669, #10b981) !important; border: none !important; }
+.gradio-container { background: rgba(17, 25, 40, 0.8) !important; backdrop-filter: blur(12px); border-radius: 20px; border: 1px solid rgba(255, 255, 255, 0.1); box-shadow: 0 8px 32px 0 rgba(0, 0, 0, 0.37); }
+#title-block { text-align: center; padding: 20px; }
+.gr-button-primary { background: linear-gradient(135deg, #059669, #10b981) !important; border: none !important; font-weight: bold !important; transition: all 0.3s ease !important; }
+.gr-button-primary:hover { transform: scale(1.02); box-shadow: 0 0 15px rgba(16, 185, 129, 0.4); }
 """
 
 with gr.Blocks(css=custom_css, theme=gr.themes.Soft()) as demo:
-    # Remplacement de gr.Div par gr.Column (Fix AttributeError)
-    with gr.Column(elem_id="title-container"):
+    with gr.Column(elem_id="title-block"):
         gr.HTML("""
-            <h1 style='color:#facc15; font-size: 2.5rem; margin:0;'>🤖 ROBOTSMALI</h1>
-            <p style='color:#94a3b8;'>Intelligence Artificielle pour le Bambara</p>
+            <h1 style='color:#facc15; font-size: 2.5rem; margin-bottom:0;'>🤖 ROBOTSMALI</h1>
+            <p style='color:#94a3b8; font-size: 1.1rem;'>Intelligence Artificielle pour la Langue Bambara</p>
             <div style="height: 3px; width: 60px; background: #facc15; margin: 15px auto;"></div>
         """)
 
     with gr.Row():
         with gr.Column():
-            gr.Markdown("### 📥 Entrée")
+            gr.Markdown("### 📥 Source Vidéo")
             v_in = gr.Video(label=None, mirror_webcam=False)
-            m_sel = gr.Dropdown(list(MODELS.keys()), value="Soloba V1 (CTC)", label="Modèle")
-            btn = gr.Button("🚀 GÉNÉRER", variant="primary")
+            m_sel = gr.Dropdown(list(MODELS.keys()), value="Soloba V1 (CTC)", label="Modèle IA")
+            btn = gr.Button("🚀 GÉNÉRER LES SOUS-TITRES", variant="primary")
             
         with gr.Column():
             gr.Markdown("### 📤 Résultat")
-            status = gr.Markdown("*Prêt...*")
+            status = gr.Markdown("*En attente de traitement...*")
             v_out = gr.Video(label=None)
 
-    # Section des exemples (Utilise maintenant MARALINKE.mp4)
+    # Section Exemples
     gr.Examples(
         examples=VIDEO_EXAMPLES,
         inputs=[v_in, m_sel],
-        label="📺 Exemples de Clips"
+        label="📺 Sélectionner une vidéo d'exemple"
     )
 
-    gr.HTML("<div style='text-align: center; color: #475569; margin-top: 30px;'>© 2025 RobotsMali - Bamako</div>")
+    gr.HTML("<div style='text-align: center; color: #475569; margin-top: 30px; font-size: 0.9rem;'>© 2025 RobotsMali • Bamako, Mali</div>")
     
     btn.click(pipeline, [v_in, m_sel], [status, v_out])
 
 if __name__ == "__main__":
-    # Petit check de debug pour le dossier examples
-    if not os.path.exists("examples/MARALINKE.mp4"):
-        print("⚠️ ATTENTION : examples/MARALINKE.mp4 est introuvable sur le serveur.")
     demo.launch(share=True, debug=True)