Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,605 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
|
| 3 |
+
# ============================================================
|
| 4 |
+
# IMPORTS & CONFIG
|
| 5 |
+
# ============================================================
|
| 6 |
+
import os, io, json, re, uuid, time, shutil, traceback, gc, asyncio
|
| 7 |
+
from moviepy.editor import *
|
| 8 |
+
import moviepy.video.fx.all as vfx
|
| 9 |
+
from moviepy.video.compositing.concatenate import concatenate_videoclips
|
| 10 |
+
from gtts import gTTS
|
| 11 |
+
from PIL import Image, ImageDraw, ImageFont, ImageFilter, ImageOps
|
| 12 |
+
import gradio as gr
|
| 13 |
+
from pydub import AudioSegment
|
| 14 |
+
from rich import print as rprint
|
| 15 |
+
import edge_tts
|
| 16 |
+
|
| 17 |
+
ROOT = os.getcwd()
|
| 18 |
+
OUT_DIR = os.path.join(ROOT, "export")
|
| 19 |
+
TMP_DIR = os.path.join(ROOT, "_tmp_capsules")
|
| 20 |
+
os.makedirs(OUT_DIR, exist_ok=True)
|
| 21 |
+
os.makedirs(TMP_DIR, exist_ok=True)
|
| 22 |
+
|
| 23 |
+
FONT_REG = "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf"
|
| 24 |
+
FONT_BOLD = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"
|
| 25 |
+
|
| 26 |
+
W, H = 1920, 1080
|
| 27 |
+
MARGIN_X, SAFE_Y_TOP = 140, 140
|
| 28 |
+
|
| 29 |
+
capsules = []
|
| 30 |
+
manifest_path = os.path.join(OUT_DIR, "manifest.json")
|
| 31 |
+
|
| 32 |
+
themes = {
|
| 33 |
+
"Bleu Professionnel": {"primary": (0, 82, 147), "secondary": (0, 126, 200)},
|
| 34 |
+
"Vert Gouvernemental": {"primary": (0, 104, 55), "secondary": (0, 155, 119)},
|
| 35 |
+
"Violet Élégant": {"primary": (74, 20, 140), "secondary": (103, 58, 183)},
|
| 36 |
+
}
|
| 37 |
+
|
| 38 |
+
# ============================================================
|
| 39 |
+
# OUTILS GÉNÉRAUX
|
| 40 |
+
# ============================================================
|
| 41 |
+
def wrap_text(text, font, max_width, draw):
|
| 42 |
+
lines = []
|
| 43 |
+
for para in text.split("\n"):
|
| 44 |
+
current = []
|
| 45 |
+
for word in para.split(" "):
|
| 46 |
+
test = " ".join(current + [word])
|
| 47 |
+
try:
|
| 48 |
+
w = draw.textlength(test, font=font)
|
| 49 |
+
except AttributeError:
|
| 50 |
+
bbox = draw.textbbox((0, 0), test, font=font)
|
| 51 |
+
w = bbox[2] - bbox[0]
|
| 52 |
+
if w <= max_width or not current:
|
| 53 |
+
current.append(word)
|
| 54 |
+
else:
|
| 55 |
+
lines.append(" ".join(current))
|
| 56 |
+
current = [word]
|
| 57 |
+
if current:
|
| 58 |
+
lines.append(" ".join(current))
|
| 59 |
+
return lines
|
| 60 |
+
|
| 61 |
+
def draw_text_shadow(draw, xy, text, font, fill=(255,255,255)):
|
| 62 |
+
x, y = xy
|
| 63 |
+
draw.text((x+2, y+2), text, font=font, fill=(0,0,0))
|
| 64 |
+
draw.text((x, y), text, font=font, fill=fill)
|
| 65 |
+
|
| 66 |
+
def safe_name(stem, ext=".mp4"):
|
| 67 |
+
stem = re.sub(r"[^\w\-]+", "_", stem)[:40]
|
| 68 |
+
return f"{stem}_{uuid.uuid4().hex[:6]}{ext}"
|
| 69 |
+
|
| 70 |
+
def save_manifest():
|
| 71 |
+
with open(manifest_path, "w", encoding="utf-8") as f:
|
| 72 |
+
json.dump({"capsules": capsules}, f, ensure_ascii=False, indent=2)
|
| 73 |
+
|
| 74 |
+
if os.path.exists(manifest_path):
|
| 75 |
+
try:
|
| 76 |
+
data = json.load(open(manifest_path, "r", encoding="utf-8"))
|
| 77 |
+
if isinstance(data, dict) and "capsules" in data:
|
| 78 |
+
capsules = data["capsules"]
|
| 79 |
+
except Exception as e:
|
| 80 |
+
rprint(f"[yellow]Impossible de charger le manifest: {e}[/yellow]")
|
| 81 |
+
|
| 82 |
+
# ============================================================
|
| 83 |
+
# 🧠 SYNTHÈSE VOCALE – gTTS / Edge / espeak / Kokoro (Hugging Face)
|
| 84 |
+
# ============================================================
|
| 85 |
+
!pip install -q transformers soundfile onnxruntime > /dev/null
|
| 86 |
+
|
| 87 |
+
from transformers import pipeline
|
| 88 |
+
import soundfile as sf
|
| 89 |
+
import asyncio
|
| 90 |
+
from pydub import AudioSegment
|
| 91 |
+
|
| 92 |
+
def tts_gtts(text, lang="fr"):
|
| 93 |
+
"""🔹 Génère avec Google TTS (rapide, en ligne)"""
|
| 94 |
+
out = os.path.join(TMP_DIR, f"gtts_{uuid.uuid4().hex}.mp3")
|
| 95 |
+
from gtts import gTTS
|
| 96 |
+
gTTS(text=text, lang=lang).save(out)
|
| 97 |
+
return out
|
| 98 |
+
|
| 99 |
+
def tts_espeak(text, voice="fr+f3"):
|
| 100 |
+
"""🔹 Génère avec eSpeak (offline, basique)"""
|
| 101 |
+
out = os.path.join(TMP_DIR, f"espeak_{uuid.uuid4().hex}.wav")
|
| 102 |
+
os.system(f'espeak-ng -v {voice} -s 165 -p 50 --stdout "{text}" > "{out}"')
|
| 103 |
+
return out
|
| 104 |
+
|
| 105 |
+
async def _tts_edge_async(text, voice="fr-FR-DeniseNeural"):
|
| 106 |
+
"""🔹 Edge-TTS (Azure Cloud)"""
|
| 107 |
+
import edge_tts
|
| 108 |
+
out = os.path.join(TMP_DIR, f"edge_{uuid.uuid4().hex}.mp3")
|
| 109 |
+
communicate = edge_tts.Communicate(text, voice)
|
| 110 |
+
await communicate.save(out)
|
| 111 |
+
return out
|
| 112 |
+
|
| 113 |
+
def tts_edge(text, voice="fr-FR-DeniseNeural"):
|
| 114 |
+
"""Génère via Edge-TTS puis convertit en WAV"""
|
| 115 |
+
try:
|
| 116 |
+
mp3_path = asyncio.run(_tts_edge_async(text, voice))
|
| 117 |
+
wav_path = os.path.join(TMP_DIR, f"edge_{uuid.uuid4().hex}.wav")
|
| 118 |
+
AudioSegment.from_file(mp3_path).export(wav_path, format="wav")
|
| 119 |
+
os.remove(mp3_path)
|
| 120 |
+
return wav_path
|
| 121 |
+
except Exception as e:
|
| 122 |
+
rprint(f"[red]Erreur Edge-TTS: {e}, fallback espeak[/red]")
|
| 123 |
+
return tts_espeak(text)
|
| 124 |
+
|
| 125 |
+
def tts_kokoro(text, langue="fr"):
|
| 126 |
+
"""🔹 Synthèse Kokoro Hugging Face (offline, voix très naturelle)"""
|
| 127 |
+
try:
|
| 128 |
+
kokoro = pipeline("text-to-speech", model="onnx-community/Kokoro-82M-v1.0-ONNX")
|
| 129 |
+
output = kokoro(text)
|
| 130 |
+
wav_path = os.path.join(TMP_DIR, f"kokoro_{uuid.uuid4().hex}.wav")
|
| 131 |
+
sf.write(wav_path, output["audio"], output["sampling_rate"])
|
| 132 |
+
return wav_path
|
| 133 |
+
except Exception as e:
|
| 134 |
+
rprint(f"[red]Erreur Kokoro: {e}, fallback espeak[/red]")
|
| 135 |
+
return tts_espeak(text)
|
| 136 |
+
|
| 137 |
+
def synth_voice(text, engine="Kokoro", gender_hint="Féminine", langue="fr"):
|
| 138 |
+
"""
|
| 139 |
+
Sélectionne automatiquement le moteur TTS selon le choix utilisateur.
|
| 140 |
+
"""
|
| 141 |
+
try:
|
| 142 |
+
if engine == "Kokoro":
|
| 143 |
+
return tts_kokoro(text, langue=langue)
|
| 144 |
+
|
| 145 |
+
elif engine == "gTTS":
|
| 146 |
+
path = tts_gtts(text, lang=langue)
|
| 147 |
+
# Ajustement léger masculin
|
| 148 |
+
if gender_hint.lower().startswith("m"):
|
| 149 |
+
snd = AudioSegment.from_file(path)
|
| 150 |
+
snd = snd._spawn(snd.raw_data, overrides={"frame_rate": int(snd.frame_rate * 0.94)}).set_frame_rate(snd.frame_rate)
|
| 151 |
+
snd.export(path, format="mp3")
|
| 152 |
+
return path
|
| 153 |
+
|
| 154 |
+
elif engine == "edge-tts":
|
| 155 |
+
voices = {
|
| 156 |
+
"fr": {"f": "fr-FR-DeniseNeural", "m": "fr-FR-AlainNeural"},
|
| 157 |
+
"nl": {"f": "nl-NL-ColetteNeural", "m": "nl-BE-ArnaudNeural"},
|
| 158 |
+
"en": {"f": "en-GB-LibbyNeural", "m": "en-GB-RyanNeural"},
|
| 159 |
+
}
|
| 160 |
+
v = voices.get(langue, voices["fr"])["f" if gender_hint.lower().startswith("f") else "m"]
|
| 161 |
+
return tts_edge(text, voice=v)
|
| 162 |
+
|
| 163 |
+
else: # espeak fallback
|
| 164 |
+
voice = f"{langue}+f3" if gender_hint.lower().startswith("f") else f"{langue}+m3"
|
| 165 |
+
return tts_espeak(text, voice=voice)
|
| 166 |
+
|
| 167 |
+
except Exception as e:
|
| 168 |
+
rprint(f"[red]Erreur synthèse {engine}: {e}, fallback espeak[/red]")
|
| 169 |
+
voice = f"{langue}+f3" if gender_hint.lower().startswith("f") else f"{langue}+m3"
|
| 170 |
+
return tts_espeak(text, voice=voice)
|
| 171 |
+
|
| 172 |
+
|
| 173 |
+
# ============================================================
|
| 174 |
+
# AUDIO NORMALISATION WAV
|
| 175 |
+
# ============================================================
|
| 176 |
+
def _normalize_audio_to_wav(in_path: str) -> str:
|
| 177 |
+
# Convertit n'importe quel format (mp3/wav) en WAV standard
|
| 178 |
+
wav_path = os.path.join(TMP_DIR, f"norm_{uuid.uuid4().hex}.wav")
|
| 179 |
+
snd = AudioSegment.from_file(in_path)
|
| 180 |
+
snd = snd.set_frame_rate(44100).set_channels(2).set_sample_width(2)
|
| 181 |
+
snd.export(wav_path, format="wav")
|
| 182 |
+
return wav_path
|
| 183 |
+
|
| 184 |
+
# ============================================================
|
| 185 |
+
# SOUS-TITRES
|
| 186 |
+
# ============================================================
|
| 187 |
+
def write_srt(text, duration):
|
| 188 |
+
parts = re.split(r'(?<=[\.!?])\s+', text.strip())
|
| 189 |
+
parts = [p for p in parts if p]
|
| 190 |
+
total = len("".join(parts)) or 1
|
| 191 |
+
cur = 0.0
|
| 192 |
+
srt = []
|
| 193 |
+
for i, p in enumerate(parts, 1):
|
| 194 |
+
prop = len(p)/total
|
| 195 |
+
start = cur
|
| 196 |
+
end = min(duration, cur + duration*prop)
|
| 197 |
+
cur = end
|
| 198 |
+
def ts(t):
|
| 199 |
+
m, s = divmod(t, 60)
|
| 200 |
+
h, m = divmod(m, 60)
|
| 201 |
+
return f"{int(h):02}:{int(m):02}:{int(s):02},000"
|
| 202 |
+
srt += [f"{i}", f"{ts(start)} --> {ts(end)}", p, ""]
|
| 203 |
+
path = os.path.join(OUT_DIR, f"srt_{uuid.uuid4().hex[:6]}.srt")
|
| 204 |
+
open(path, "w", encoding="utf-8").write("\n".join(srt))
|
| 205 |
+
return path
|
| 206 |
+
|
| 207 |
+
# ============================================================
|
| 208 |
+
# CRÉATION FOND
|
| 209 |
+
# ============================================================
|
| 210 |
+
def make_background(titre, sous_titre, texte_ecran, theme, logo_path, logo_pos, img_fond, fond_mode="plein écran"):
|
| 211 |
+
c = themes[theme]
|
| 212 |
+
bg = Image.new("RGB", (W, H), c["primary"])
|
| 213 |
+
if img_fond and os.path.exists(img_fond):
|
| 214 |
+
img = Image.open(img_fond).convert("RGB")
|
| 215 |
+
if fond_mode == "plein écran":
|
| 216 |
+
img = img.resize((W, H))
|
| 217 |
+
img = img.filter(ImageFilter.GaussianBlur(1))
|
| 218 |
+
overlay = Image.new("RGBA", (W, H), (*c["primary"], 90))
|
| 219 |
+
bg = Image.alpha_composite(img.convert("RGBA"), overlay).convert("RGB")
|
| 220 |
+
elif fond_mode == "moitié gauche":
|
| 221 |
+
img = img.resize((W//2, H))
|
| 222 |
+
mask = Image.linear_gradient("L").resize((W//2, H))
|
| 223 |
+
color = Image.new("RGB", (W//2, H), c["primary"])
|
| 224 |
+
comp = Image.composite(img, color, ImageOps.invert(mask))
|
| 225 |
+
bg.paste(comp, (0, 0))
|
| 226 |
+
elif fond_mode == "moitié droite":
|
| 227 |
+
img = img.resize((W//2, H))
|
| 228 |
+
mask = Image.linear_gradient("L").resize((W//2, H))
|
| 229 |
+
color = Image.new("RGB", (W//2, H), c["primary"])
|
| 230 |
+
comp = Image.composite(color, img, mask)
|
| 231 |
+
bg.paste(comp, (W//2, 0))
|
| 232 |
+
elif fond_mode == "moitié bas":
|
| 233 |
+
img = img.resize((W, H//2))
|
| 234 |
+
mask = Image.linear_gradient("L").rotate(90).resize((W, H//2))
|
| 235 |
+
color = Image.new("RGB", (W, H//2), c["primary"])
|
| 236 |
+
comp = Image.composite(color, img, mask)
|
| 237 |
+
bg.paste(comp, (0, H//2))
|
| 238 |
+
|
| 239 |
+
draw = ImageDraw.Draw(bg)
|
| 240 |
+
f_title = ImageFont.truetype(FONT_BOLD, 84)
|
| 241 |
+
f_sub = ImageFont.truetype(FONT_REG, 44)
|
| 242 |
+
f_text = ImageFont.truetype(FONT_REG, 40)
|
| 243 |
+
f_small = ImageFont.truetype(FONT_REG, 30)
|
| 244 |
+
draw.rectangle([(0, 0), (W, 96)], fill=c["secondary"])
|
| 245 |
+
draw.rectangle([(0, H-96), (W, H)], fill=c["secondary"])
|
| 246 |
+
draw_text_shadow(draw, (MARGIN_X, 30), "CPAS BRUXELLES • SERVICE PUBLIC", f_small)
|
| 247 |
+
draw_text_shadow(draw, (W//2-280, H-72), "📞 0800 35 550 • 🌐 cpasbru.irisnet.be", f_small)
|
| 248 |
+
draw_text_shadow(draw, (MARGIN_X, SAFE_Y_TOP), titre, f_title)
|
| 249 |
+
draw_text_shadow(draw, (MARGIN_X, SAFE_Y_TOP+100), sous_titre, f_sub)
|
| 250 |
+
y = SAFE_Y_TOP + 200
|
| 251 |
+
for line in texte_ecran.split("\n"):
|
| 252 |
+
for l in wrap_text("• "+line.strip("• "), f_text, W-MARGIN_X*2, draw):
|
| 253 |
+
draw_text_shadow(draw, (MARGIN_X, y), l, f_text)
|
| 254 |
+
y += 55
|
| 255 |
+
if logo_path and os.path.exists(logo_path):
|
| 256 |
+
logo = Image.open(logo_path).convert("RGBA")
|
| 257 |
+
logo.thumbnail((260,260))
|
| 258 |
+
lw, lh = logo.size
|
| 259 |
+
pos = (50,50) if logo_pos=="haut-gauche" else (W-lw-50,50) if logo_pos=="haut-droite" else ((W-lw)//2,50)
|
| 260 |
+
bg.paste(logo, pos, logo)
|
| 261 |
+
out = os.path.join(TMP_DIR, f"fond_{uuid.uuid4().hex[:6]}.png")
|
| 262 |
+
bg.save(out)
|
| 263 |
+
return out
|
| 264 |
+
|
| 265 |
+
# ============================================================
|
| 266 |
+
# FFMPEG FALLBACK (sans temp_audiofile)
|
| 267 |
+
# ============================================================
|
| 268 |
+
def _write_video_with_fallback(final_clip, out_path_base, fps=24):
|
| 269 |
+
attempts = [
|
| 270 |
+
{"ext": ".mp4", "codec": "libx264", "audio_codec": "aac"},
|
| 271 |
+
{"ext": ".mp4", "codec": "mpeg4", "audio_codec": "aac"},
|
| 272 |
+
{"ext": ".mp4", "codec": "libx264", "audio_codec": "libmp3lame"},
|
| 273 |
+
]
|
| 274 |
+
ffmpeg_common = [
|
| 275 |
+
"-pix_fmt", "yuv420p",
|
| 276 |
+
"-movflags", "+faststart",
|
| 277 |
+
"-threads", "1",
|
| 278 |
+
"-max_muxing_queue_size", "1024",
|
| 279 |
+
"-shortest"
|
| 280 |
+
]
|
| 281 |
+
last_err = None
|
| 282 |
+
for i, opt in enumerate(attempts, 1):
|
| 283 |
+
out = out_path_base if out_path_base.endswith(opt["ext"]) else out_path_base + opt["ext"]
|
| 284 |
+
try:
|
| 285 |
+
rprint(f"[cyan]FFmpeg try #{i}: codec={opt['codec']} audio={opt['audio_codec']} -> {out}[/cyan]")
|
| 286 |
+
final_clip.write_videofile(
|
| 287 |
+
out,
|
| 288 |
+
fps=fps,
|
| 289 |
+
codec=opt["codec"],
|
| 290 |
+
audio_codec=opt["audio_codec"],
|
| 291 |
+
audio=True,
|
| 292 |
+
ffmpeg_params=ffmpeg_common,
|
| 293 |
+
logger=None,
|
| 294 |
+
threads=1,
|
| 295 |
+
)
|
| 296 |
+
if os.path.exists(out) and os.path.getsize(out) > 150000:
|
| 297 |
+
return out
|
| 298 |
+
except Exception as e:
|
| 299 |
+
last_err = f"{type(e).__name__}: {e}\n" + traceback.format_exc()
|
| 300 |
+
rprint(f"[yellow]Échec essai #{i}: {last_err}[/yellow]")
|
| 301 |
+
raise RuntimeError(last_err or "Échec inconnu de l'encodage FFmpeg")
|
| 302 |
+
|
| 303 |
+
def generate_sadtalker_video(image_path, audio_path, output_dir=TMP_DIR, fps=25):
|
| 304 |
+
"""
|
| 305 |
+
Génère une vidéo animée du visage à partir d'une image et d'un fichier audio,
|
| 306 |
+
en utilisant SadTalker.
|
| 307 |
+
"""
|
| 308 |
+
try:
|
| 309 |
+
os.makedirs(output_dir, exist_ok=True)
|
| 310 |
+
out_path = os.path.join(output_dir, f"sadtalker_{uuid.uuid4().hex[:6]}.mp4")
|
| 311 |
+
cmd = (
|
| 312 |
+
f'cd SadTalker && python inference.py '
|
| 313 |
+
f'--driven_audio "{audio_path}" '
|
| 314 |
+
f'--source_image "{image_path}" '
|
| 315 |
+
f'--result_dir "{output_dir}" '
|
| 316 |
+
f'--still --enhancer gfpgan --fps {fps}'
|
| 317 |
+
)
|
| 318 |
+
os.system(cmd)
|
| 319 |
+
|
| 320 |
+
# Cherche le dernier fichier mp4 généré
|
| 321 |
+
candidates = [os.path.join(output_dir, f) for f in os.listdir(output_dir) if f.endswith(".mp4")]
|
| 322 |
+
if not candidates:
|
| 323 |
+
rprint("[red]❌ Aucune sortie SadTalker générée[/red]")
|
| 324 |
+
return None
|
| 325 |
+
latest = max(candidates, key=os.path.getctime)
|
| 326 |
+
os.rename(latest, out_path)
|
| 327 |
+
if os.path.exists(out_path):
|
| 328 |
+
return out_path
|
| 329 |
+
return None
|
| 330 |
+
except Exception as e:
|
| 331 |
+
rprint(f"[red]Erreur SadTalker: {e}[/red]")
|
| 332 |
+
return None
|
| 333 |
+
|
| 334 |
+
#=============================================================
|
| 335 |
+
# ============================================================
|
| 336 |
+
# BUILD CAPSULE – Version SadTalker (qualité type Sora 2)
|
| 337 |
+
# ============================================================
|
| 338 |
+
def build_capsule(titre, sous_titre, texte_voix, texte_ecran, theme,
|
| 339 |
+
image_fond=None, logo_path=None, logo_pos="haut-gauche",
|
| 340 |
+
fond_mode="plein écran",
|
| 341 |
+
video_presentateur=None, voix_type="Féminine",
|
| 342 |
+
position_presentateur="bottom-right", plein=False,
|
| 343 |
+
moteur_voix="Edge-TTS (voix naturelle)", langue="fr"):
|
| 344 |
+
"""
|
| 345 |
+
Construit une capsule vidéo complète avec :
|
| 346 |
+
- TTS Edge/gTTS/espeak
|
| 347 |
+
- Génération visuelle (fond + texte)
|
| 348 |
+
- Présentateur animé via SadTalker
|
| 349 |
+
"""
|
| 350 |
+
try:
|
| 351 |
+
# ====================================================
|
| 352 |
+
# 1️⃣ Synthèse vocale
|
| 353 |
+
# ====================================================
|
| 354 |
+
# Sélection du moteur vocal selon le choix utilisateur
|
| 355 |
+
if moteur_voix.startswith("Kokoro"):
|
| 356 |
+
engine = "Kokoro"
|
| 357 |
+
elif moteur_voix.startswith("Edge"):
|
| 358 |
+
engine = "edge-tts"
|
| 359 |
+
elif moteur_voix.startswith("gTTS"):
|
| 360 |
+
engine = "gTTS"
|
| 361 |
+
else:
|
| 362 |
+
engine = "espeak"
|
| 363 |
+
|
| 364 |
+
audio_path_mp = synth_voice(texte_voix, engine=engine, gender_hint=voix_type, langue=langue)
|
| 365 |
+
audio_path = _normalize_audio_to_wav(audio_path_mp)
|
| 366 |
+
|
| 367 |
+
if not os.path.exists(audio_path) or os.path.getsize(audio_path) < 1000:
|
| 368 |
+
raise RuntimeError(f"Audio non valide : {audio_path}")
|
| 369 |
+
|
| 370 |
+
audio = AudioFileClip(audio_path)
|
| 371 |
+
dur = float(audio.duration or 5.0)
|
| 372 |
+
target_fps = 25 # ✅ FPS fixe haute qualité
|
| 373 |
+
|
| 374 |
+
# ====================================================
|
| 375 |
+
# 2️⃣ Génération du fond graphique
|
| 376 |
+
# ====================================================
|
| 377 |
+
fond_path = make_background(
|
| 378 |
+
titre, sous_titre, texte_ecran, theme,
|
| 379 |
+
logo_path, logo_pos, image_fond, fond_mode
|
| 380 |
+
)
|
| 381 |
+
bg = ImageClip(fond_path).set_duration(dur)
|
| 382 |
+
|
| 383 |
+
# ====================================================
|
| 384 |
+
# 3️⃣ Génération présentateur (SadTalker)
|
| 385 |
+
# ====================================================
|
| 386 |
+
clips = [bg]
|
| 387 |
+
if video_presentateur and os.path.exists(video_presentateur):
|
| 388 |
+
ext = os.path.splitext(video_presentateur)[1].lower()
|
| 389 |
+
video_path = None
|
| 390 |
+
|
| 391 |
+
if ext in [".jpg", ".jpeg", ".png"]:
|
| 392 |
+
rprint("[cyan]🎭 Génération visage animé avec SadTalker...[/cyan]")
|
| 393 |
+
synced = generate_sadtalker_video(video_presentateur, audio_path, fps=target_fps)
|
| 394 |
+
if synced:
|
| 395 |
+
video_path = synced
|
| 396 |
+
rprint("[green]✅ SadTalker : visage animé généré[/green]")
|
| 397 |
+
else:
|
| 398 |
+
rprint("[red]⚠️ SadTalker n’a pas pu produire de vidéo[/red]")
|
| 399 |
+
else:
|
| 400 |
+
rprint("[yellow]⚠️ SadTalker attend une image (portrait), pas une vidéo.[/yellow]")
|
| 401 |
+
video_path = video_presentateur
|
| 402 |
+
|
| 403 |
+
if video_path and os.path.exists(video_path):
|
| 404 |
+
v = VideoFileClip(video_path).without_audio()
|
| 405 |
+
v = v.fx(vfx.loop, duration=dur)
|
| 406 |
+
|
| 407 |
+
if plein:
|
| 408 |
+
v = v.resize((W, H))
|
| 409 |
+
else:
|
| 410 |
+
v = v.resize(width=480)
|
| 411 |
+
pos_map = {
|
| 412 |
+
"bottom-right": ("right", "bottom"),
|
| 413 |
+
"bottom-left": ("left", "bottom"),
|
| 414 |
+
"top-right": ("right", "top"),
|
| 415 |
+
"top-left": ("left", "top"),
|
| 416 |
+
"center": ("center", "center"),
|
| 417 |
+
}
|
| 418 |
+
v = v.set_position(pos_map.get(position_presentateur, ("right", "bottom")))
|
| 419 |
+
|
| 420 |
+
clips.append(v)
|
| 421 |
+
|
| 422 |
+
# ====================================================
|
| 423 |
+
# 4️⃣ Composition finale et export
|
| 424 |
+
# ====================================================
|
| 425 |
+
final = CompositeVideoClip(clips).set_audio(audio.set_fps(44100))
|
| 426 |
+
name = safe_name(f"{titre}_{langue}")
|
| 427 |
+
out_base = os.path.join(OUT_DIR, name)
|
| 428 |
+
out = _write_video_with_fallback(final, out_base, fps=target_fps)
|
| 429 |
+
|
| 430 |
+
# ====================================================
|
| 431 |
+
# 5️⃣ Sous-titres + Manifest
|
| 432 |
+
# ====================================================
|
| 433 |
+
srt_path = write_srt(texte_voix, dur)
|
| 434 |
+
capsules.append({
|
| 435 |
+
"file": out,
|
| 436 |
+
"title": titre,
|
| 437 |
+
"langue": langue,
|
| 438 |
+
"voice": voix_type,
|
| 439 |
+
"theme": theme,
|
| 440 |
+
"duration": round(dur, 1)
|
| 441 |
+
})
|
| 442 |
+
save_manifest()
|
| 443 |
+
|
| 444 |
+
# ====================================================
|
| 445 |
+
# 6️⃣ Nettoyage
|
| 446 |
+
# ====================================================
|
| 447 |
+
audio.close()
|
| 448 |
+
final.close()
|
| 449 |
+
bg.close()
|
| 450 |
+
try:
|
| 451 |
+
if os.path.exists(audio_path_mp): os.remove(audio_path_mp)
|
| 452 |
+
if os.path.exists(audio_path): os.remove(audio_path)
|
| 453 |
+
except:
|
| 454 |
+
pass
|
| 455 |
+
gc.collect()
|
| 456 |
+
|
| 457 |
+
return out, f"✅ Capsule {langue.upper()} créée ({dur:.1f}s, voix {voix_type})", srt_path
|
| 458 |
+
|
| 459 |
+
except Exception as e:
|
| 460 |
+
err_msg = f"❌ Erreur: {e}\n\nTraceback:\n{traceback.format_exc()}"
|
| 461 |
+
rprint(f"[red]{err_msg}[/red]")
|
| 462 |
+
return None, err_msg, None
|
| 463 |
+
|
| 464 |
+
|
| 465 |
+
# ============================================================
|
| 466 |
+
# TABLEAU, ASSEMBLAGE ET GESTION DES CAPSULES
|
| 467 |
+
# ============================================================
|
| 468 |
+
def table_capsules():
|
| 469 |
+
return [[i+1, c["title"], c.get("langue","fr").upper(),
|
| 470 |
+
f"{c['duration']}s", c["theme"], c["voice"],
|
| 471 |
+
os.path.basename(c["file"])]
|
| 472 |
+
for i, c in enumerate(capsules)]
|
| 473 |
+
|
| 474 |
+
|
| 475 |
+
def assemble_final():
|
| 476 |
+
if not capsules:
|
| 477 |
+
return None, "❌ Aucune capsule."
|
| 478 |
+
clips = [VideoFileClip(c["file"]) for c in capsules]
|
| 479 |
+
final = concatenate_videoclips(clips, method="compose")
|
| 480 |
+
try:
|
| 481 |
+
out = _write_video_with_fallback(final, os.path.join(OUT_DIR, safe_name("VIDEO_COMPLETE")), fps=25)
|
| 482 |
+
return out, f"🎉 Vidéo finale prête ({len(capsules)} capsules)."
|
| 483 |
+
finally:
|
| 484 |
+
for c in clips:
|
| 485 |
+
try: c.close()
|
| 486 |
+
except: pass
|
| 487 |
+
try: final.close()
|
| 488 |
+
except: pass
|
| 489 |
+
|
| 490 |
+
|
| 491 |
+
def supprimer_capsule(index):
|
| 492 |
+
try:
|
| 493 |
+
idx = int(index) - 1
|
| 494 |
+
if 0 <= idx < len(capsules):
|
| 495 |
+
fichier = capsules[idx]["file"]
|
| 496 |
+
if os.path.exists(fichier):
|
| 497 |
+
os.remove(fichier)
|
| 498 |
+
del capsules[idx]
|
| 499 |
+
save_manifest()
|
| 500 |
+
return f"🗑 Capsule supprimée : {fichier}", table_capsules()
|
| 501 |
+
else:
|
| 502 |
+
return "⚠️ Index invalide.", table_capsules()
|
| 503 |
+
except Exception as e:
|
| 504 |
+
return f"❌ Erreur lors de la suppression : {e}", table_capsules()
|
| 505 |
+
|
| 506 |
+
|
| 507 |
+
def deplacer_capsule(index, direction):
|
| 508 |
+
try:
|
| 509 |
+
idx = int(index) - 1
|
| 510 |
+
if direction == "up" and idx > 0:
|
| 511 |
+
capsules[idx - 1], capsules[idx] = capsules[idx], capsules[idx - 1]
|
| 512 |
+
elif direction == "down" and idx < len(capsules) - 1:
|
| 513 |
+
capsules[idx + 1], capsules[idx] = capsules[idx], capsules[idx + 1]
|
| 514 |
+
save_manifest()
|
| 515 |
+
return f"🔁 Capsule déplacée {direction}.", table_capsules()
|
| 516 |
+
except Exception as e:
|
| 517 |
+
return f"❌ Erreur de déplacement : {e}", table_capsules()
|
| 518 |
+
|
| 519 |
+
|
| 520 |
+
# ============================================================
|
| 521 |
+
# INTERFACE GRADIO (adaptée pour SadTalker)
|
| 522 |
+
# ============================================================
|
| 523 |
+
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
| 524 |
+
gr.Markdown("## 🎬 Créateur de Capsules CPAS – Version SadTalker (réaliste)")
|
| 525 |
+
|
| 526 |
+
with gr.Tab("Créer une capsule"):
|
| 527 |
+
with gr.Row():
|
| 528 |
+
with gr.Column():
|
| 529 |
+
image_fond = gr.Image(label="🖼 Image de fond", type="filepath")
|
| 530 |
+
fond_mode = gr.Radio(["plein écran", "moitié gauche", "moitié droite", "moitié bas"],
|
| 531 |
+
label="Mode d'affichage du fond", value="plein écran")
|
| 532 |
+
logo_path = gr.Image(label="🏛 Logo", type="filepath")
|
| 533 |
+
logo_pos = gr.Radio(["haut-gauche","haut-droite","centre"],
|
| 534 |
+
label="Position logo", value="haut-gauche")
|
| 535 |
+
# SadTalker prend une IMAGE, pas une vidéo
|
| 536 |
+
video_presentateur = gr.Image(label="🧑🎨 Image du présentateur (SadTalker)", type="filepath")
|
| 537 |
+
position_presentateur = gr.Radio(["bottom-right","bottom-left","top-right","top-left","center"],
|
| 538 |
+
label="Position", value="bottom-right")
|
| 539 |
+
plein = gr.Checkbox(label="Plein écran présentateur", value=False)
|
| 540 |
+
with gr.Column():
|
| 541 |
+
titre = gr.Textbox(label="Titre", value="Aide médicale urgente / Dringende medische hulp")
|
| 542 |
+
sous_titre = gr.Textbox(label="Sous-titre", value="Soins accessibles à tous / Toegankelijke zorg voor iedereen")
|
| 543 |
+
theme = gr.Radio(list(themes.keys()), label="Thème", value="Bleu Professionnel")
|
| 544 |
+
langue = gr.Radio(["fr","nl"], label="Langue de la voix", value="fr")
|
| 545 |
+
voix_type = gr.Radio(["Féminine","Masculine"], label="Voix IA", value="Féminine")
|
| 546 |
+
moteur_voix = gr.Radio(
|
| 547 |
+
["Kokoro (HuggingFace, offline)", "Edge-TTS (voix naturelle)", "gTTS (en ligne)", "espeak-ng (offline)"],
|
| 548 |
+
label="Moteur voix",
|
| 549 |
+
value="Kokoro (HuggingFace, offline)"
|
| 550 |
+
)
|
| 551 |
+
|
| 552 |
+
|
| 553 |
+
texte_voix = gr.Textbox(label="Texte voix off", lines=4,
|
| 554 |
+
value="Bonjour, le CPAS de Bruxelles vous aide pour vos soins de santé.")
|
| 555 |
+
texte_ecran = gr.Textbox(label="Texte à l'écran", lines=4,
|
| 556 |
+
value="💊 Aides médicales\n🏥 Soins urgents\n📋 Formalités simplifiées")
|
| 557 |
+
btn = gr.Button("🎬 Créer Capsule", variant="primary")
|
| 558 |
+
|
| 559 |
+
sortie = gr.Video(label="Capsule générée")
|
| 560 |
+
srt_out = gr.File(label="Sous-titres .srt")
|
| 561 |
+
statut = gr.Markdown()
|
| 562 |
+
|
| 563 |
+
# Onglet gestion
|
| 564 |
+
with gr.Tab("Gestion & Assemblage"):
|
| 565 |
+
gr.Markdown("### 🗂 Gestion des capsules")
|
| 566 |
+
liste = gr.Dataframe(
|
| 567 |
+
headers=["N°","Titre","Langue","Durée","Thème","Voix","Fichier"],
|
| 568 |
+
value=table_capsules(),
|
| 569 |
+
interactive=False
|
| 570 |
+
)
|
| 571 |
+
with gr.Row():
|
| 572 |
+
index = gr.Number(label="Index capsule", value=1, precision=0)
|
| 573 |
+
btn_up = gr.Button("⬆️ Monter")
|
| 574 |
+
btn_down = gr.Button("⬇️ Descendre")
|
| 575 |
+
btn_del = gr.Button("🗑 Supprimer")
|
| 576 |
+
message = gr.Markdown()
|
| 577 |
+
|
| 578 |
+
btn_up.click(lambda i: deplacer_capsule(i, "up"), [index], [message, liste])
|
| 579 |
+
btn_down.click(lambda i: deplacer_capsule(i, "down"), [index], [message, liste])
|
| 580 |
+
btn_del.click(supprimer_capsule, [index], [message, liste])
|
| 581 |
+
|
| 582 |
+
gr.Markdown("### 🎬 Assemblage final")
|
| 583 |
+
btn_asm = gr.Button("🎥 Assembler la vidéo complète", variant="primary")
|
| 584 |
+
sortie_finale = gr.Video(label="Vidéo finale")
|
| 585 |
+
btn_asm.click(lambda: assemble_final(), [], [sortie_finale, message])
|
| 586 |
+
|
| 587 |
+
# Fonction principale pour Gradio
|
| 588 |
+
def creer_capsule_ui(t, st, tv, te, th, img, fmode, logo, pos_logo, vp, vx, pos_p, plein, motor, langue):
|
| 589 |
+
vid, msg, srt = build_capsule(t, st, tv, te, th,
|
| 590 |
+
img, logo, pos_logo, fmode,
|
| 591 |
+
vp, vx, pos_p, plein,
|
| 592 |
+
motor, langue)
|
| 593 |
+
return vid, srt, msg, table_capsules()
|
| 594 |
+
|
| 595 |
+
btn.click(
|
| 596 |
+
creer_capsule_ui,
|
| 597 |
+
[titre, sous_titre, texte_voix, texte_ecran, theme,
|
| 598 |
+
image_fond, fond_mode, logo_path, logo_pos,
|
| 599 |
+
video_presentateur, voix_type, position_presentateur,
|
| 600 |
+
plein, moteur_voix, langue],
|
| 601 |
+
[sortie, srt_out, statut, liste]
|
| 602 |
+
)
|
| 603 |
+
|
| 604 |
+
print("🚀 Lancement de l'interface SadTalker FR/NL…")
|
| 605 |
+
demo.launch(share=True, debug=True)
|