changes

app.py

@@ -1,3 +1,4 @@
+#app.py
 import gradio as gr
 import os
 import shutil
@@ -44,72 +45,74 @@ def process_video(
     user_music: Optional[str] = None,
     show_progress_bar: bool = True,
 ):
-    """Build the final video using user‑defined visual parameters (brightness, contrast, gamma)."""
+    """Build the final video with a single encoding pass."""
 
     if not accumulated_videos:
         raise ValueError("❌ Please upload at least one background video (.mp4) before generating.")
 
     approx_words = int(target_duration * WORDS_PER_SECOND)
 
-    # --- 1. Script (AI or custom) ---
+    # ── 1. Script (AI ou perso) ──────────────────────────────
     if script_mode == "Use my script":
         if not custom_script or not custom_script.strip():
             raise ValueError("❌ You selected 'Use my script' but the script field is empty!")
         script = custom_script.strip()
-        title = generate_title(script)
-        description = generate_description(script)
     else:
         prompt = (
-            f"You are a video creation expert. Here is the context: {context.strip()}\n"
-            f"Instruction: {instruction.strip()}\n"
-            f"🔴 Strict target duration: {target_duration}s — ≈ {approx_words} words (must be respected)."
+            f"You are a video creation expert.\n\nContext:\n{context.strip()}\n\n"
+            f"Instruction:\n{instruction.strip()}\n\n"
+            f"🔴 Strict target duration: {target_duration}s — ≈ {approx_words} words."
         )
         script = generate_script(prompt)
-        title = generate_title(script)
-        description = generate_description(script)
 
-    # --- 2. Prepare folders ---
-    for folder in ("./assets/audio", "./assets/backgrounds", "./assets/output"):
+    title       = generate_title(script)
+    description = generate_description(script)
+
+    # ── 2. Préparation répertoires ───────────────────────────
+    for folder in ("./assets/audio", "./assets/backgrounds", "./assets/output", "./assets/video_music"):
         os.makedirs(folder, exist_ok=True)
 
-    voice_path = "./assets/audio/voice.mp3"
-    final_no_subs = "./assets/output/final_video.mp4"
-    final_with_subs = "./assets/output/final_video_subtitles.mp4"
+    voice_path     = "./assets/audio/voice.mp3"
+    final_no_subs  = "./assets/output/final_video.mp4"
 
-    # --- 3. Copy videos ---
-    for f in os.listdir("./assets/backgrounds"):
-        if f.lower().endswith(".mp4"):
-            os.remove(os.path.join("./assets/backgrounds", f))
+    # ── 3. Copie unique des vidéos de fond ───────────────────
     for idx, v in enumerate(accumulated_videos):
         if not os.path.isfile(v) or not v.lower().endswith(".mp4"):
             raise ValueError(f"❌ Invalid file: {v}")
         safe_copy(v, os.path.join("./assets/backgrounds", f"video_{idx:03d}.mp4"))
 
-    # --- 4. AI voice ---
-    generate_voice(script, voice_path)
+    # ── 4. Voix IA (cache disque) ────────────────────────────
+    if not os.path.isfile(voice_path):
+        generate_voice(script, voice_path)
 
-    # --- 5. Video montage ---
-    music_path = user_music if user_music and os.path.isfile(user_music) else None
+    # ── 5. Montage silencieux (pas d’audio) ──────────────────
     _, out_no_audio = get_video_montage_from_folder(
         folder_path="./assets/backgrounds",
-        audio_path=voice_path,
+        audio_path=voice_path,                 # juste pour la durée, pas d’injection
         output_dir="./assets/video_music",
-        lum=lum,
-        contrast=contrast,
-        gamma=gamma,
+        lum=lum, contrast=contrast, gamma=gamma,
         show_progress_bar=show_progress_bar,
     )
 
-    # --- 6. Mixing & subtitles ---
-    edit_video(out_no_audio, voice_path, music_path, final_no_subs)
-
+    # ── 6. Sous-titres (optionnel) ───────────────────────────
+    subs = None
     if add_subs:
         segments = transcribe_audio_to_subs(voice_path)
-        subs = chunk_text_by_words(segments, max_words=3)
-        add_subtitles_to_video(final_no_subs, subs, final_with_subs)
-        return script, title, description, final_with_subs
-    else:
-        return script, title, description, final_no_subs
+        subs     = chunk_text_by_words(segments, max_words=3)
+
+    # ── 7. Mux final en une passe ────────────────────────────
+    music_path = user_music if user_music and os.path.isfile(user_music) else None
+
+    edit_video(
+        video_path   = out_no_audio,
+        audio_path   = voice_path,
+        music_path   = music_path,
+        output_path  = final_no_subs,
+        music_volume = 0.10,
+        subtitles    = subs,      # ← injectés ici
+    )
+
+    return script, title, description, final_no_subs
 
 # ──────────────────────────────────────────────────────────────────────────────
 # Upload helper

scripts/edit_video.py

@@ -1,69 +1,79 @@
-# ============================
-# get_footage.py  (unchanged)
-# ============================
+from pathlib import Path
+from typing import Optional, List, Dict
 
-# (contenu identique à la précédente version – pas de modification)
+from moviepy import (
+    VideoFileClip, AudioFileClip, CompositeAudioClip,
+    CompositeVideoClip
+)
 
-# ============================
-# edit_video.py (révision => musique optionnelle et volume paramétrable)
-# ============================
-
-"""Assemble la voix IA et, si fourni, la musique de fond.
-
-Appel :
-    edit_video(
-        video_path="./assets/video_music/video_silent.mp4",
-        audio_path="./assets/audio/voice.mp3",
-        music_path=None,          # ou chemin .mp3 / .wav
-        output_path="./assets/output/final_video.mp4",
-        music_volume=0.10,        # volume musique (0‑1)
-    )
-"""
-
-from moviepy import VideoFileClip, AudioFileClip, CompositeAudioClip
-import os
+from scripts.generate_subtitles import create_animated_subtitle_clip
 
 
 def edit_video(
     video_path: str,
     audio_path: str,
-    music_path: str | None,
+    music_path: Optional[str],
     output_path: str,
     *,
     music_volume: float = 0.10,
+    subtitles: Optional[List[Dict]] = None,   # ← nouveau
 ):
-    video_clip = VideoFileClip(video_path)
+    """
+    Encodage final : ajoute voix, (optionnel) musique et sous-titres
+    en UNE seule passe.
+    """
+    vid_clip   = VideoFileClip(video_path)
     voice_clip = AudioFileClip(audio_path)
-    tracks = [voice_clip]
 
-    if music_path and os.path.isfile(music_path):
+    # ── piste audio composite ─────────────────────────────────
+    tracks = [voice_clip]
+    if music_path and Path(music_path).is_file():
         try:
             music_clip = (
                 AudioFileClip(music_path)
                 .with_volume_scaled(music_volume)
-                .with_duration(video_clip.duration)
+                .with_duration(vid_clip.duration)
             )
             tracks.insert(0, music_clip)
         except Exception as err:
-            print(f"⚠️ Musique ignorée : {err}")
+            print(f"⚠️ Music ignored: {err}")
+
+    final_audio = CompositeAudioClip(tracks).with_duration(vid_clip.duration)
 
-    final_audio = CompositeAudioClip(tracks).with_duration(video_clip.duration)
-    final_clip = video_clip.with_audio(final_audio)
+    # ── couche(s) vidéo / sous-titres ─────────────────────────
+    layers = [vid_clip]
+    if subtitles:
+        w, h = vid_clip.size
+        for sub in subtitles:
+            layers.append(
+                create_animated_subtitle_clip(
+                    sub["text"], sub["start"], sub["end"], w, h
+                )
+            )
+
+    final_clip = (
+        CompositeVideoClip(layers, size=vid_clip.size)
+        .with_duration(vid_clip.duration)
+        .with_audio(final_audio)
+    )
 
+    # ── export ────────────────────────────────────────────────
+    Path(output_path).parent.mkdir(parents=True, exist_ok=True)
     final_clip.write_videofile(
         output_path,
         codec="libx264",
         audio_codec="aac",
         fps=30,
-        threads=4,
+        threads=os.cpu_count(),
         preset="medium",
-        ffmpeg_params=["-pix_fmt", "yuv420p"],
+        ffmpeg_params=["-pix_fmt", "yuv420p"]
     )
-    print(f"✅ Vidéo générée : {output_path}")
+    print(f"✅ Video written → {output_path}")
 
-    video_clip.close()
+    # ── nettoyage ─────────────────────────────────────────────
     voice_clip.close()
     if "music_clip" in locals():
         music_clip.close()
     final_audio.close()
     final_clip.close()
+    vid_clip.close()

scripts/generate_scripts.py

@@ -1,3 +1,4 @@
+#generate_scripts.py
 import os
 import re
 import json

scripts/generate_subtitles.py

@@ -25,7 +25,8 @@ SUBTITLE_COLORS = [
     "white", "yellow", "cyan", "deeppink", "gold", "lightgreen", "magenta", "orange"
 ]
 
-
+WHISPER = whisper.load_model("medium",
+                             device="cuda" if torch.cuda.is_available() else "cpu")
 
 
 def color_for_word(word: str) -> str:
@@ -85,29 +86,14 @@ def save_subtitles_to_srt(subtitles, output_path):
 
 
 @spaces.GPU()
-def transcribe_audio_to_subs(audio_path):
-    """
-    Transcrit le fichier audio en texte (via Whisper), retourne la liste
-    des segments start/end/text, et sauvegarde en .srt.
-    """
-    print("🎙️ Transcription avec Whisper...")
-    model = whisper.load_model("medium", device="cuda" if torch.cuda.is_available() else "cpu")
-    result = model.transcribe(audio_path)
-
-    subtitles = [{
-        "start": seg['start'],
-        "end": seg['end'],
-        "text": seg['text']
-    } for seg in result['segments']]
-
-    print(f"📝 {len(subtitles)} sous-titres générés.")
-
-    # Sauvegarde .srt
-    base_name = os.path.splitext(audio_path)[0]
-    srt_path = f"{base_name}.srt"
-    save_subtitles_to_srt(subtitles, srt_path)
-    print(f"💾 Sous-titres enregistrés dans : {srt_path}")
-
+def transcribe_audio_to_subs(audio_path: str):
+    """Transcrit audio → segments (pas de re-load du modèle)."""
+    print("🎙️ Transcription Whisper (cached model)...")
+    result = WHISPER.transcribe(audio_path)
+    subtitles = [
+        {"start": s["start"], "end": s["end"], "text": s["text"]}
+        for s in result["segments"]
+    ]
     return subtitles
 
 def format_subtitle_text(text, max_chars=50):
@@ -231,5 +217,4 @@ def add_subtitles_to_video(video_path, subtitles, output_file="./assets/output/v
         ffmpeg_params=["-pix_fmt", "yuv420p"]
     )
 
-    print(f"✅ Vidéo Shorts/TikTok prête : {output_file}")
-
+    print(f"✅ Vidéo Shorts/TikTok prête : {output_file}")

scripts/get_footage.py

@@ -23,7 +23,6 @@ from moviepy import (
     CompositeVideoClip
 )
 
-FONT_PATH = "C:/Windows/Fonts/arialbd.ttf"
 
 def add_pan_effect(clip):
     """
@@ -130,7 +129,7 @@ def apply_crossfade_effects(clips, duration=0.12):
 
 def get_video_montage_from_folder(
     folder_path: str = "./assets/videos",
-    audio_path: str = "./assets/audio/voice.mp3",
+    audio_path: str | None = None,          # ← devient optionnel
     output_dir: str = "./assets/backgrounds",
     lum: float = 6.0,
     contrast: float = 1.0,
@@ -138,103 +137,61 @@ def get_video_montage_from_folder(
     show_progress_bar: bool = True,
 ):
     """
-    1) Parcourt tous les fichiers vidéo dans 'folder_path'
-    2) Construit un montage vertical (1080x1920) en appliquant dynamic_effect()
-       et un crossfade entre chaque clip.
-    3) La durée totale est bornée à la durée de l'audio (on coupe le surplus).
-    4) Exporte deux versions : avec et sans audio.
+    Construit un montage vertical 1080×1920 SANS piste audio.
+    Si audio_path est fourni, on s’en sert seulement pour borner la durée.
+    Retourne (None, path_video_silent).
     """
 
-    # Prépare les chemins de sortie
     os.makedirs(output_dir, exist_ok=True)
-    output_with_audio = os.path.join(output_dir, "video_with_audio.mp4")
-    output_no_audio   = os.path.join(output_dir, "video_silent.mp4")
+    output_no_audio = os.path.join(output_dir, "video_silent.mp4")
 
-    # Charge l'audio pour connaître la durée cible
-    voiceover = AudioFileClip(audio_path)
-    audio_duration = voiceover.duration
-    print(f"🎧 Durée audio : {audio_duration:.2f} s")
+    # Durée cible = longueur voice-over (facultatif)
+    audio_duration = None
+    if audio_path and os.path.isfile(audio_path):
+        audio_duration = AudioFileClip(audio_path).duration
+        print(f"🎧 Target duration (voice-over) : {audio_duration:.2f} s")
 
-    # Liste de tous les fichiers vidéo dans le dossier
+    # Récupère toutes les vidéos du dossier
     all_videos = [
         f for f in os.listdir(folder_path)
         if f.lower().endswith((".mp4", ".mov", ".avi", ".mkv"))
     ]
-
     if not all_videos:
-        print(f"❌ Aucune vidéo trouvée dans le dossier : {folder_path}")
-        return None, None
-
-    clips = []
-    total_duration = 0.0
-
-    # Parcours des vidéos dans l'ordre
-    for idx, video_file in enumerate(all_videos):
-        video_path = os.path.join(folder_path, video_file)
-        try:
-            clip = VideoFileClip(video_path)
-
-            # Redimensionne en 1080x1920 (vertical)
-            target_w, target_h = 1080, 1920
-            clip_ar = clip.w / clip.h
-            target_ar = target_w / target_h
-
-            if clip_ar > target_ar:
-                # On adapte la hauteur
-                clip = clip.resized(height=target_h)
-                # On coupe la largeur
-                clip = clip.cropped(width=target_w, x_center=clip.w / 2)
-            else:
-                # On adapte la largeur
-                clip = clip.resized(width=target_w)
-                # On coupe la hauteur
-                clip = clip.cropped(height=target_h, y_center=clip.h / 2)
-
-            # Applique l’effet dynamique
-            dynamic_clip = dynamic_effect(clip, lum, contrast, gamma)
-            clips.append(dynamic_clip)
-            total_duration += dynamic_clip.duration
-
-            # Si la somme dépasse la durée audio, on arrête la boucle
-            if total_duration >= audio_duration:
-                break
-
-        except Exception as e:
-            print(f"⚠️ Erreur avec le fichier {video_file} : {e}")
+        raise RuntimeError(f"❌ No videos found in {folder_path}")
+
+    clips, total_duration = [], 0.0
+
+    for video_file in all_videos:
+        clip = VideoFileClip(os.path.join(folder_path, video_file))
+
+        # Resize/crop en 1080×1920
+        target_w, target_h = 1080, 1920
+        clip_ar, target_ar = clip.w / clip.h, target_w / target_h
+        if clip_ar > target_ar:
+            clip = clip.resized(height=target_h).cropped(width=target_w, x_center=clip.w/2)
+        else:
+            clip = clip.resized(width=target_w).cropped(height=target_h, y_center=clip.h/2)
+
+        # Effets dynamiques
+        dynamic_clip = dynamic_effect(clip, lum, contrast, gamma)
+        clips.append(dynamic_clip)
+        total_duration += dynamic_clip.duration
+
+        if audio_duration and total_duration >= audio_duration:
+            break
 
     if not clips:
-        print("❌ Aucun clip valide. Montage impossible.")
-        return None, None
+        raise RuntimeError("❌ Montage impossible (no valid clips).")
 
-    # Crossfade entre les clips
     clips = apply_crossfade_effects(clips, duration=0.15)
+    final_clip = concatenate_videoclips(clips, method="compose")
+    if audio_duration:
+        final_clip = final_clip.subclipped(0, audio_duration)
 
-    # Concaténation, borne la durée totale à celle de l'audio
-    final_clip = concatenate_videoclips(clips, method="compose").subclipped(0, audio_duration)
-
-    # Overlay (par ex. barre de progression)
     if show_progress_bar:
         final_clip = add_timer_overlay(final_clip)
 
-    # --------------------
-    # 1) Version AVEC audio
-    # --------------------
-    final_clip_with_audio = final_clip.with_audio(voiceover)
-    final_clip_with_audio.write_videofile(
-        output_with_audio,
-        codec='libx264',
-        audio_codec='aac',
-        fps=30,
-        threads=4,
-        preset="medium",
-        ffmpeg_params=["-pix_fmt", "yuv420p"]
-    )
-    print(f"✅ Montage créé (AVEC audio) : {output_with_audio}")
-
-
-    # --------------------
-    # 2) Version SANS audio
-    # --------------------
+    # Export silencieux
     final_clip.write_videofile(
         output_no_audio,
         codec='libx264',
@@ -246,14 +203,11 @@ def get_video_montage_from_folder(
     )
     print(f"✅ Montage créé (SANS audio) : {output_no_audio}")
 
-    # Libère la mémoire
+    # Clean
     for c in clips:
         c.close()
-    voiceover.close()
     final_clip.close()
-    final_clip_with_audio.close()
-
 
-    return output_with_audio, output_no_audio
+    return None, output_no_audio