AzALlN/whisperx-pt-asr

Temporarily disable research_metrics to fix runtime error

Fix import path for research_metrics module

dfdf

ok

fdf

ok

feat: Add comprehensive research analytics module - Add research_metrics.py with scientific comparison capabilities - Include system performance metrics (CPU, GPU, memory usage) - Add transcription quality analysis (confidence stats, timing precision) - Include speaker diarization analysis and pronunciation capabilities assessment - Add model information and cost/scalability metrics - Integrate research_analytics field in JSON response for comparative studies

feat: Add word-level confidence and timestamps to JSON structure - Add word-level confidence scores and precise timestamps - Add words array to segments with detailed word data - Enhanced JSON for pronunciation assessment capabilities

projeto 100% funcional

fix: adicionar api_name explícito para endpoint /process_audio

fix: eliminar recursão infinita - usar WhisperX diretamente no Space

feat: usar HF_API_KEY automaticamente da variável secreta do Space

fix: melhorar tratamento de erros e logging no app.py

Fix: Change from gr.Interface to gr.Blocks for better API control - Replace gr.Interface with gr.Blocks - Fix API endpoint issues causing timeouts - Better Gradio API support

Fix: Add complete service.py content - Complete WhisperX service module with all functions - Fix import error for transcribe_audio_file function

Refactor: WhisperX app now returns structured JSON - Add service.py module for modular API access - Refactor app.py to use service and return JSON - Simple interface with structured data output

fdfd

Fix: Corrigir requirements.txt e habilitar show_error na API - Renomear requirements.tx para requirements.txt - Adicionar show_error=True no demo.launch() para debug - Configurar HF_API_KEY automático - Corrigir formato requirements.txt com --extra-index-url

rttr

dffdfd

dff

fdffd

fddfdf

fdddf

ffdfdfd

fddffddf

dfdffd

fd

dfd

dffddf

ffgfg

🔧 Align WhisperX implementation with official documentation

🛡️ Fix WhisperX VAD configuration and cuDNN error handling

🗣️ Fix language detection - set Portuguese as default in model loading

🛡️ Add robust cuDNN error handling with automatic CPU fallback

🔧 Melhorar tratamento de erros na função transcribe_audio

🎮 Forçar GPU por padrão no HF Spaces + diagnóstico melhorado - Detecta HF Spaces via SPACE_ID - Força CUDA por padrão no HF Spaces - Adiciona logs detalhados de GPU/CPU - Fallback inteligente se CUDA falhar

🔧 Habilitar show_error=True para debug da API

🚀 Otimizações para HF Spaces: GPU por padrão, only large-v3, português padrão - Configuração automática de GPU (CUDA) quando disponível - Modelo fixo large-v3 para máxima qualidade - Português como idioma padrão - Interface simplificada - Otimizações de performance para HF Spaces

🔄 Restore WhisperX with speaker diarization - local processing with full speaker detection

🔧 Fix port configuration for HF Spaces - use default port

🚀 Initial clean commit: HF API version with Gradio 5.32.0 & requests 2.32.3