Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -108,6 +108,26 @@ def keyword_score(text: str, custom_keywords: str = "", weight_emotion: float =
|
|
| 108 |
score += 0.0005 * len(text)
|
| 109 |
return score
|
| 110 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 111 |
def llm_rank_segments(candidates: List[Segment], num_segments: int, custom_instructions: str = "") -> List[Segment]:
|
| 112 |
"""Ask the LLM to pick segments based on criteria."""
|
| 113 |
if not LLM_AVAILABLE:
|
|
@@ -140,9 +160,39 @@ def llm_rank_segments(candidates: List[Segment], num_segments: int, custom_instr
|
|
| 140 |
return candidates[:num_segments]
|
| 141 |
|
| 142 |
def select_segments(transcript_txt: str, use_llm: bool, num_segments: int,
|
| 143 |
-
custom_keywords: str,
|
| 144 |
weight_emotion: float, weight_break: float,
|
| 145 |
weight_learn: float, weight_viral: float) -> List[Segment]:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 146 |
segs = parse_transcript(transcript_txt)
|
| 147 |
if not segs:
|
| 148 |
raise ValueError("Nenhum trecho válido encontrado na transcrição.")
|
|
@@ -154,7 +204,7 @@ def select_segments(transcript_txt: str, use_llm: bool, num_segments: int,
|
|
| 154 |
top = segs[:min(20, len(segs))]
|
| 155 |
|
| 156 |
if use_llm and LLM_AVAILABLE:
|
| 157 |
-
ranked = llm_rank_segments(top, num_segments,
|
| 158 |
return ranked
|
| 159 |
|
| 160 |
return top[:num_segments]
|
|
@@ -266,16 +316,26 @@ def edit_sequence_with_segments(tree: ET.ElementTree, segs: List[Segment]) -> ET
|
|
| 266 |
|
| 267 |
# ---- Gradio app ----
|
| 268 |
def process_xml_and_transcript(premiere_xml_file, transcript_txt_file, use_llm,
|
| 269 |
-
num_segments, custom_keywords,
|
| 270 |
weight_emotion, weight_break, weight_learn, weight_viral):
|
| 271 |
-
if premiere_xml_file is None
|
| 272 |
-
return "Envie o XML do Premiere
|
|
|
|
|
|
|
|
|
|
| 273 |
|
| 274 |
-
|
| 275 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 276 |
|
| 277 |
segs = select_segments(transcript, use_llm and LLM_AVAILABLE, num_segments,
|
| 278 |
-
custom_keywords,
|
| 279 |
weight_emotion, weight_break, weight_learn, weight_viral)
|
| 280 |
|
| 281 |
tree = ET.parse(premiere_xml_file.name)
|
|
@@ -285,13 +345,16 @@ def process_xml_and_transcript(premiere_xml_file, transcript_txt_file, use_llm,
|
|
| 285 |
out_path = os.path.join(OUTPUT_DIR, f"{base}_EDITADO.xml")
|
| 286 |
tree.write(out_path, encoding="utf-8", xml_declaration=True)
|
| 287 |
|
| 288 |
-
|
|
|
|
| 289 |
for i, s in enumerate(segs, 1):
|
| 290 |
dur_sec = (s.end_f - s.start_f) / FPS
|
| 291 |
resumo += f"{i}. {s.start_tc} → {s.end_tc} ({dur_sec:.1f}s)\n"
|
| 292 |
-
|
|
|
|
|
|
|
| 293 |
|
| 294 |
-
status = f"✓ LLM disponível: {LLM_AVAILABLE} | LLM usado: {use_llm and LLM_AVAILABLE}"
|
| 295 |
return resumo, out_path, status
|
| 296 |
|
| 297 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
@@ -301,7 +364,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
| 301 |
with gr.Row():
|
| 302 |
with gr.Column():
|
| 303 |
xml_in = gr.File(label="📁 XML da sequência (FCP XML)", file_types=[".xml"])
|
| 304 |
-
txt_in = gr.File(label="📄 Transcrição (.txt)
|
| 305 |
|
| 306 |
with gr.Column():
|
| 307 |
gr.Markdown("### ⚙️ Configurações")
|
|
@@ -312,23 +375,29 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
| 312 |
)
|
| 313 |
num_segments = gr.Slider(
|
| 314 |
minimum=2, maximum=10, step=1, value=5,
|
| 315 |
-
label="📊 Número de segmentos
|
| 316 |
info="Quantos trechos incluir no vídeo final"
|
| 317 |
)
|
| 318 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 319 |
with gr.Accordion("🎯 Palavras-chave Personalizadas", open=False):
|
| 320 |
custom_keywords = gr.Textbox(
|
| 321 |
label="Adicione palavras-chave importantes (separadas por vírgula)",
|
| 322 |
placeholder="Exemplo: transformação, resultado, método, estratégia",
|
| 323 |
-
info="Trechos com essas palavras terão prioridade máxima (peso 3.0)"
|
| 324 |
-
)
|
| 325 |
-
|
| 326 |
-
with gr.Accordion("📝 Instruções em Texto Livre para o LLM", open=False):
|
| 327 |
-
custom_instructions = gr.Textbox(
|
| 328 |
-
label="Instruções adicionais para o LLM",
|
| 329 |
-
placeholder="Exemplo: Prefira trechos que mostrem resultados concretos e evite introduções longas",
|
| 330 |
-
lines=3,
|
| 331 |
-
info="Só funciona se o LLM estiver ativado"
|
| 332 |
)
|
| 333 |
|
| 334 |
with gr.Accordion("⚖️ Ajuste Fino dos Pesos de Pontuação", open=False):
|
|
@@ -354,16 +423,31 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
| 354 |
run_btn.click(
|
| 355 |
process_xml_and_transcript,
|
| 356 |
inputs=[xml_in, txt_in, use_llm, num_segments, custom_keywords,
|
| 357 |
-
|
| 358 |
outputs=[resumo_out, file_out, status_out]
|
| 359 |
)
|
| 360 |
|
| 361 |
gr.Markdown("""
|
| 362 |
-
### 💡
|
| 363 |
-
|
| 364 |
-
|
| 365 |
-
-
|
| 366 |
-
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 367 |
""")
|
| 368 |
|
| 369 |
if __name__ == "__main__":
|
|
|
|
| 108 |
score += 0.0005 * len(text)
|
| 109 |
return score
|
| 110 |
|
| 111 |
+
def parse_manual_timecodes(manual_input: str) -> List[tuple]:
|
| 112 |
+
"""Parse manual timecode ranges from user input.
|
| 113 |
+
Expected format: hh:mm:ss:ff - hh:mm:ss:ff (one per line or comma-separated)
|
| 114 |
+
Returns list of (start_tc, end_tc) tuples
|
| 115 |
+
"""
|
| 116 |
+
manual_ranges = []
|
| 117 |
+
# Replace commas with newlines for flexibility
|
| 118 |
+
normalized = manual_input.replace(",", "\n")
|
| 119 |
+
lines = [l.strip() for l in normalized.splitlines() if l.strip()]
|
| 120 |
+
|
| 121 |
+
pat = re.compile(r"(\d{2}:\d{2}:\d{2}[:;]\d{2})\s*[-–—]\s*(\d{2}:\d{2}:\d{2}[:;]\d{2})")
|
| 122 |
+
|
| 123 |
+
for line in lines:
|
| 124 |
+
m = pat.search(line)
|
| 125 |
+
if m:
|
| 126 |
+
start_tc, end_tc = m.groups()
|
| 127 |
+
manual_ranges.append((start_tc, end_tc))
|
| 128 |
+
|
| 129 |
+
return manual_ranges
|
| 130 |
+
|
| 131 |
def llm_rank_segments(candidates: List[Segment], num_segments: int, custom_instructions: str = "") -> List[Segment]:
|
| 132 |
"""Ask the LLM to pick segments based on criteria."""
|
| 133 |
if not LLM_AVAILABLE:
|
|
|
|
| 160 |
return candidates[:num_segments]
|
| 161 |
|
| 162 |
def select_segments(transcript_txt: str, use_llm: bool, num_segments: int,
|
| 163 |
+
custom_keywords: str, manual_timecodes: str,
|
| 164 |
weight_emotion: float, weight_break: float,
|
| 165 |
weight_learn: float, weight_viral: float) -> List[Segment]:
|
| 166 |
+
|
| 167 |
+
# Check if user provided manual timecodes
|
| 168 |
+
manual_ranges = parse_manual_timecodes(manual_timecodes)
|
| 169 |
+
|
| 170 |
+
if manual_ranges:
|
| 171 |
+
# Manual mode: use only the timecodes provided by user
|
| 172 |
+
result_segs = []
|
| 173 |
+
for start_tc, end_tc in manual_ranges:
|
| 174 |
+
try:
|
| 175 |
+
start_f = parse_timecode_to_frames(start_tc)
|
| 176 |
+
end_f = parse_timecode_to_frames(end_tc)
|
| 177 |
+
if end_f > start_f:
|
| 178 |
+
result_segs.append(Segment(
|
| 179 |
+
start_tc=start_tc,
|
| 180 |
+
end_tc=end_tc,
|
| 181 |
+
start_f=start_f,
|
| 182 |
+
end_f=end_f,
|
| 183 |
+
text=f"Corte manual {start_tc} - {end_tc}",
|
| 184 |
+
score=100.0
|
| 185 |
+
))
|
| 186 |
+
except Exception as e:
|
| 187 |
+
print(f"Erro ao processar timecode manual {start_tc}-{end_tc}: {e}")
|
| 188 |
+
continue
|
| 189 |
+
|
| 190 |
+
if not result_segs:
|
| 191 |
+
raise ValueError("Nenhum timecode manual válido encontrado.")
|
| 192 |
+
|
| 193 |
+
return result_segs
|
| 194 |
+
|
| 195 |
+
# Automatic mode: use transcript + scoring
|
| 196 |
segs = parse_transcript(transcript_txt)
|
| 197 |
if not segs:
|
| 198 |
raise ValueError("Nenhum trecho válido encontrado na transcrição.")
|
|
|
|
| 204 |
top = segs[:min(20, len(segs))]
|
| 205 |
|
| 206 |
if use_llm and LLM_AVAILABLE:
|
| 207 |
+
ranked = llm_rank_segments(top, num_segments, "")
|
| 208 |
return ranked
|
| 209 |
|
| 210 |
return top[:num_segments]
|
|
|
|
| 316 |
|
| 317 |
# ---- Gradio app ----
|
| 318 |
def process_xml_and_transcript(premiere_xml_file, transcript_txt_file, use_llm,
|
| 319 |
+
num_segments, custom_keywords, manual_timecodes,
|
| 320 |
weight_emotion, weight_break, weight_learn, weight_viral):
|
| 321 |
+
if premiere_xml_file is None:
|
| 322 |
+
return "Envie o XML do Premiere.", None, f"LLM disponível: {LLM_AVAILABLE}"
|
| 323 |
+
|
| 324 |
+
# Check if manual timecodes were provided
|
| 325 |
+
manual_ranges = parse_manual_timecodes(manual_timecodes)
|
| 326 |
|
| 327 |
+
if manual_ranges:
|
| 328 |
+
# Manual mode: don't need transcript
|
| 329 |
+
transcript = ""
|
| 330 |
+
else:
|
| 331 |
+
# Automatic mode: need transcript
|
| 332 |
+
if transcript_txt_file is None:
|
| 333 |
+
return "Envie a transcrição em .txt ou forneça minutagens manuais.", None, f"LLM disponível: {LLM_AVAILABLE}"
|
| 334 |
+
with open(transcript_txt_file.name, "r", encoding="utf-8") as f:
|
| 335 |
+
transcript = f.read()
|
| 336 |
|
| 337 |
segs = select_segments(transcript, use_llm and LLM_AVAILABLE, num_segments,
|
| 338 |
+
custom_keywords, manual_timecodes,
|
| 339 |
weight_emotion, weight_break, weight_learn, weight_viral)
|
| 340 |
|
| 341 |
tree = ET.parse(premiere_xml_file.name)
|
|
|
|
| 345 |
out_path = os.path.join(OUTPUT_DIR, f"{base}_EDITADO.xml")
|
| 346 |
tree.write(out_path, encoding="utf-8", xml_declaration=True)
|
| 347 |
|
| 348 |
+
mode = "MANUAL" if manual_ranges else "AUTOMÁTICO"
|
| 349 |
+
resumo = f"✂️ {len(segs)} cortes aplicados - Modo: {mode} (24 fps):\n\n"
|
| 350 |
for i, s in enumerate(segs, 1):
|
| 351 |
dur_sec = (s.end_f - s.start_f) / FPS
|
| 352 |
resumo += f"{i}. {s.start_tc} → {s.end_tc} ({dur_sec:.1f}s)\n"
|
| 353 |
+
if not manual_ranges:
|
| 354 |
+
resumo += f" Score: {s.score:.1f} | {s.text[:150]}\n"
|
| 355 |
+
resumo += "\n"
|
| 356 |
|
| 357 |
+
status = f"✓ Modo: {mode} | LLM disponível: {LLM_AVAILABLE} | LLM usado: {use_llm and LLM_AVAILABLE and not manual_ranges}"
|
| 358 |
return resumo, out_path, status
|
| 359 |
|
| 360 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
|
|
| 364 |
with gr.Row():
|
| 365 |
with gr.Column():
|
| 366 |
xml_in = gr.File(label="📁 XML da sequência (FCP XML)", file_types=[".xml"])
|
| 367 |
+
txt_in = gr.File(label="📄 Transcrição (.txt) - Opcional se usar minutagens manuais", file_types=[".txt"])
|
| 368 |
|
| 369 |
with gr.Column():
|
| 370 |
gr.Markdown("### ⚙️ Configurações")
|
|
|
|
| 375 |
)
|
| 376 |
num_segments = gr.Slider(
|
| 377 |
minimum=2, maximum=10, step=1, value=5,
|
| 378 |
+
label="📊 Número de segmentos (só no modo automático)",
|
| 379 |
info="Quantos trechos incluir no vídeo final"
|
| 380 |
)
|
| 381 |
|
| 382 |
+
with gr.Accordion("✂️ MINUTAGENS MANUAIS (Sobrescreve tudo)", open=True):
|
| 383 |
+
manual_timecodes = gr.Textbox(
|
| 384 |
+
label="Cole aqui os timecodes exatos que você quer cortar",
|
| 385 |
+
placeholder="Exemplo:\n00:01:23:15 - 00:02:45:10\n00:05:30:00 - 00:07:15:22\n00:10:00:05 - 00:12:30:18",
|
| 386 |
+
lines=5,
|
| 387 |
+
info="⚠️ Se preencher este campo, o app ignora a transcrição e todos os outros parâmetros, cortando EXATAMENTE o que você especificou"
|
| 388 |
+
)
|
| 389 |
+
gr.Markdown("""
|
| 390 |
+
**Formatos aceitos:**
|
| 391 |
+
- `hh:mm:ss:ff - hh:mm:ss:ff` (um por linha)
|
| 392 |
+
- Pode separar por vírgula também
|
| 393 |
+
- Exemplo: `00:01:30:00 - 00:02:00:15, 00:05:10:00 - 00:06:20:10`
|
| 394 |
+
""")
|
| 395 |
+
|
| 396 |
with gr.Accordion("🎯 Palavras-chave Personalizadas", open=False):
|
| 397 |
custom_keywords = gr.Textbox(
|
| 398 |
label="Adicione palavras-chave importantes (separadas por vírgula)",
|
| 399 |
placeholder="Exemplo: transformação, resultado, método, estratégia",
|
| 400 |
+
info="Trechos com essas palavras terão prioridade máxima (peso 3.0) - Só funciona no modo automático"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 401 |
)
|
| 402 |
|
| 403 |
with gr.Accordion("⚖️ Ajuste Fino dos Pesos de Pontuação", open=False):
|
|
|
|
| 423 |
run_btn.click(
|
| 424 |
process_xml_and_transcript,
|
| 425 |
inputs=[xml_in, txt_in, use_llm, num_segments, custom_keywords,
|
| 426 |
+
manual_timecodes, weight_emotion, weight_break, weight_learn, weight_viral],
|
| 427 |
outputs=[resumo_out, file_out, status_out]
|
| 428 |
)
|
| 429 |
|
| 430 |
gr.Markdown("""
|
| 431 |
+
### 💡 Modos de uso:
|
| 432 |
+
|
| 433 |
+
**🎯 MODO MANUAL (Recomendado para controle total)**
|
| 434 |
+
- Preencha o campo "Minutagens Manuais" com seus timecodes exatos
|
| 435 |
+
- A transcrição se torna opcional
|
| 436 |
+
- Todos os outros parâmetros são ignorados
|
| 437 |
+
- O corte será feito EXATAMENTE como você especificou
|
| 438 |
+
|
| 439 |
+
**🤖 MODO AUTOMÁTICO**
|
| 440 |
+
- Deixe as minutagens manuais vazias
|
| 441 |
+
- Envie a transcrição com timecodes
|
| 442 |
+
- Configure LLM, palavras-chave e pesos conforme desejado
|
| 443 |
+
- O app escolhe os melhores trechos automaticamente
|
| 444 |
+
|
| 445 |
+
**Exemplos de minutagens manuais:**
|
| 446 |
+
```
|
| 447 |
+
00:01:23:15 - 00:02:45:10
|
| 448 |
+
00:05:30:00 - 00:07:15:22
|
| 449 |
+
00:10:00:05 - 00:12:30:18
|
| 450 |
+
```
|
| 451 |
""")
|
| 452 |
|
| 453 |
if __name__ == "__main__":
|