Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload README.md with huggingface_hub

Upload app.py with huggingface_hub

Upload app.py with huggingface_hub

Upload Dockerfile with huggingface_hub

Upload Dockerfile with huggingface_hub

Upload README.md with huggingface_hub

Upload app.py with huggingface_hub

BUGFIX S2ST: Auto-fallback for speaker ref in unified pipeline

BUGFIX: Auto-generate default speaker ref for XTTS (Prevent None crash)

S2ST PRO: Forced GPU Mode (No CPU Fallback) + Google Translate + Faster-Whisper

BUGFIX: Add CPU fallback for Faster-Whisper (Fix CUDA error on startup)

HOTFIX: Removed entire try/except block to purge indentation error

HOTFIX: Fixed duplicate try block in translate action

APP UPDATE: Deep-Translator Integration

DEPS FIX: Switched to deep-translator (Fix httpx conflict)

DEPS FIX: Force install faster-whisper + googletrans

HOTFIX: Removed duplicate try-block causing IndentationError

SPEED: Migrated to Google Translate API (Removed NLLB) for Instant Results

NLLB FALLBACK: Beam (Strict) -> Sampling (Creative) for stubborn translations

BUGFIX ARABIC LOOP: Corrected NLLB Code to arb_Arab + Garbage Prevention

AUDIO FIX: Disable Output Denoising (Fix Old Mic Sound) - Pure XTTS 48k

STRICT MODE: NLLB Deterministic Beam Search (x10) - Pure Translation

OPTIM: Set NLLB Temperature to 0.4 (Medium-Low) for Stability

SPEED: Migrated to Faster-Whisper (CTranslate2)

PURITY GUARD: Strict Anti-Language Mixing + Latin/Arabic Enforcer

ARABIC FIX & DOUBLE CLEAN: Input/Output Denoising + Script Check

FIX: Robust DeepFilterNet Init + Restore Denoising

OPTIM: Audio Padding (1.5s) + Strict NLLB Config

FIX: Remove broken download_model import

FIX: Monkeypatch torchaudio.backend for DeepFilterNet compat

FIX: Sentence Completion Guardrails (Auto-Ellipsis)

PRO UPGRADE: NLLB-1.3B + DeepFilterNet + Translation Fix

DEPS: Add deepfilternet for audio quality enhancement

S2ST UPDATE: Implement unified Speech-to-Speech action for better quality

PRODUCTION RELEASE: Final fixes for Translation mapping and TTS audio backend

BULLETPROOF FIX: Monkeypatch torchaudio to bypass missing torchcodec

FINAL FIX: Force torchaudio backend to bypass torchcodec error

FINAL FIX: Trigger build with torchcodec

FIX: Missing torchcodec for XTTS-v2

FIX: Robust TTS loading and build trigger

FIX: Switch to coqui-tts package

FIX: Correct NLLB model ID typo