MTxStudio - AI Video Voiceover & Multi-Track Studio

MTxStudio is a professional cross-platform application built for Windows and macOS (Apple Silicon M-Series and Intel x64), offering an end-to-end production pipeline from automated batch AI video voiceover, instant voice cloning, multi-character conversation studio, physical modeling acoustic sound studio, to hardware-accelerated multi-track video editing.

Batch AI Video Voiceover & Whisper Transcription

MTxStudio automates the complete re-voicing workflow from raw source videos:

• Multi-Source Video Ingestion: Automatic downloading via yt-dlp from social media platforms or batch local file imports (.mp4, .mov, .mkv, .avi, .webm).
• Precise Whisper STT Transcription: Powered by Faster-Whisper and WhisperX for millisecond-accurate sentence alignment with timecodes and automatic source language detection.
• Multi-Provider AI Translation: Connects OpenAI (GPT-4o), DeepSeek (V3, R1), Google Gemini (1.5 Pro, 2.0 Flash), Anthropic Claude (3.5 Sonnet), and local Ollama models.
• AI 1-Click Polish: One-click Vietnamese spell checking, punctuation correction, and natural spoken dialogue refinement.
• Visual Review Studio: Parallel timeline view of original and translated lines, direct text editing, line splitting, line merging, non-verbal expression tags insertion, and segment audio audition before approval.

OmniVoice Synthesis & Zero-Shot Voice Cloning

Next-generation audio intelligence running locally on your hardware:

• OmniVoice Diffusion LM Engine: High-fidelity speech synthesis supporting CPU bfloat16 and GPU CUDA float16 acceleration.
• Zero-Shot Voice Cloning: Requires only a 3 to 10 second audio reference to extract unique vocal acoustic embeddings (Voice ADN) and save as reusable Voice Presets.
• Multi-Attribute Voice Design: Flexibly configure voices across 5 attribute dimensions: Gender, Age, Pitch, Expression Style (whisper, standard), and over 20 regional accents and dialects.
• Voice Identity Seed Lock: Lock the generative Seed to ensure persistent, identical character timbre across episodic series.
• Non-Verbal Expression Tags: Insert physiological voice nuances directly into scripts: [laughter], [sigh], [surprise-ah], [dissatisfaction-hnn].

Multi-Character Conversation Studio

• Chat Bubble Timeline Editor: Intuitive conversation timeline interface with individual avatar representations and distinct color coding for each speaker.
• Character Cast Manager: Assemble custom AI cast profiles with assigned voices, pitch contours, speeds, and Seed locks, reusable across podcast episodes and audio dramas.
• Natural Interaction Tuning: Calibrate inter-turn conversational pauses (300ms - 800ms) and natural speech overlap thresholds.
• Versatile Audio Exporting: Export unified audio mixes, isolated per-character multi-track stems, or inject directly into video tracks with speaker name subtitle overlays.

Professional Multi-Track Video Editor

• Multi-Track Timeline: Independent management of Video Tracks, Voiceover Tracks, Background Music Tracks, Subtitle Tracks, and Overlay Image Tracks.
• Studio Editing Suite: Millisecond-accurate playhead, smart magnetic snapping, clip splitting (Ctrl + B), clip duplication (Ctrl + D), head/tail trimming, and multi-level Undo/Redo history.
• Canvas Preview & Transform Bounding Box: Real-time 2D/WebGL preview with on-canvas bounding box controls for dragging position X/Y, scaling, and rotating clips directly.
• Multiple Aspect Ratio Support: Seamlessly toggle between 16:9 (YouTube, TV), 9:16 (TikTok, Shorts, Reels), 1:1 (Square), 4:3, and 3:4.
• Auto Subtitle Editor with Per-Line Voice Override: Automatic sentence segmenting by timestamp, default voice assignment, and per-line voice overriding for distinct dialogue roles, coupled with 1-click Batch TTS synthesis.
• Integrated AI Image Generator: Enter prompts to synthesize AI artwork (DALL-E 3, Gemini Imagen, Stable Diffusion) and place directly onto timeline tracks.
• Hardware-Accelerated Export Engine: FFmpeg Complex Filtergraph pipeline exporting 720p HD, 1080p Full HD, and 4K Ultra HD videos with NVENC, Intel QuickSync, AMD AMF, and Apple Silicon VideoToolbox acceleration.

Sound Studio, Heritage Instruments & Meditation Audio

• Bamboo Flute & Xiao Studio: Full support for standard flutes (C5, D5, E5, F5, G5, A4, Bb4, Xiao F4, C4), Vietnamese Cam Am note converter (VietnameseCamAmConverter recognizing Do2, Sol2, Mi2, Fa2, Do3, lasol), automatic Transpose Tone when switching flutes, portamento glides, vibrato LFO, and authentic breath chiff noise.
• Vision Sheet Music Scanner: AI vision OCR scanning printed sheet music or camera snapshots to parse notes and Solfege into the flute and guitar editors.
• Saxophone Studio: Kenny G soprano, alto, and tenor saxophone physical modeling with scoop inflections, expressive vibrato, and nocturnal subtone warmth.
• Vietnamese Heritage Instruments: Dan Bau with portamento bending and pure harmonics, Dan Tranh with 16-string sweeps and tremolo picking, and Dan Nguyet with courtly fret-pressing dignity.
• Karplus-Strong Classical & Fingerstyle Guitar: Masterful guitar techniques including slap palm percussion, natural harmonics, tremolo picking, rasgueado strums, and chord rolls (Sunflower Paddy Sun 145 measures complete, Echoes Between the Lands).
• ChordPro & Lead Sheet Editor: ChordPro notation ([Am], [G], [C]) combined with vocal lyrics across 10 musical genres and 8 accompaniment rhythms.
• Zen Spa & Sound Therapy: Tibetan singing bowls, wind chimes, temple wooden fish (muyu), Solfeggio healing frequencies (432Hz, 528Hz), forest streams, and natural rainfall.

Silent Video Visual Intelligence (VLM)

• AI Vision Frame Analysis: Automatic keyframe extraction using FFmpeg scene cut detection, analyzed through Moondream2, Florence-2, Qwen2.5-VL, and Google Gemini Vision for scene context, character actions, and emotional atmosphere.
• Contextual Narration Scripting: Automatically synthesizes timecode-matched scripts tailored to specific presentation styles: Storytelling, Product Review, Documentary, or Instructional.
• Global Script Smoothing: Merges fragmented scene descriptions into a coherent, captivating storyline ready for Review Studio TTS synthesis.

AI Singing Engine & Screen Recorder

• AI Singing Engine: Integrates AceStep and DiffSinger services to manipulate F0 pitch curves, vocal vibrato, and syllable articulation from musical notes and lyrics, featuring automatic harmonic backing generation.
• App Showcase Screen Recorder: Sleek floating action button (FAB) for native high-definition 30 FPS full-screen capture via FFmpeg (gdigrab on Windows, avfoundation on macOS) or canvas stream chunks, recording system audio and microphone inputs simultaneously into MTxStudio_Recordings.

Local Relational Storage & Multi-Threaded Batch Rendering

• Comprehensive Source Traceability: Trace rendered video files back to raw video URLs, Whisper transcripts, AI translations, approved scripts, Voice Templates, and render timestamps.
• Organized Project Tree: Hierarchical categorization: Channel -> Category -> Project.
• Secure Local SQLite3 Storage: Persistent on-device relational database for configurations, scripts, projects, and custom pronunciation lexicons with zero cloud exposure.
• Concurrent Batch Render Queue: Decoupled multi-threaded job queue balancing CPU/GPU workloads with real-time FPS progress and ETA tracking.

Extensible Modules Ecosystem

MTxStudio features a modular architecture following strict lifecycle management (Install, Activate, Deactivate, Configure), allowing creators to activate and tailor specialized feature modules according to their workflow needs:

MTxDownloader - Video Link CRUD & Bulk Downloader

• Manage multi-platform video URL catalogs from YouTube, TikTok, Facebook Reels, Instagram Reels, Douyin, and X/Twitter.
• Support Rolling Queue bulk downloading with adaptive host resource monitoring (capping CPU and RAM utilization under 90%).
• Automatically organize and categorize download directories into standardized Channel and Category folder structures.
• Maintain full source traceability including original URL, author metadata, download date, resolution, and video duration.
• Integrated Downloader Editor Vault for quickly pushing downloaded videos into the re-voicing pipeline or timeline editor.

MTxRain - Ambient Rain & Storm Soundscape Studio

• Generate non-repeating procedural rain and storm audio waveforms running 100% on-device via physical modeling DSP algorithms.
• Acoustic impact surface matrix: rain on corrugated iron roof, window glass, tree leaves, asphalt pavement, and camping tents.
• Independent 5-layer sound mixer: Rain Drops layer, Rain Hum/Patter layer, Wind Howl layer, Thunder Strike layer, and Room Ambience layer.
• Customizable collision frequency, randomized thunder interval distance, room reverb depth, and infinite loopable audio export.

Atmospheric Visual FX Suite

• Real-time 60 FPS multi-layered particle and atmospheric visual overlay engine optimized directly for the Video Editor Canvas and Timeline.
• Rain on Glass: Realistic sliding droplets, water bead accumulation, and natural light refraction.
• Winter Snow: Parallax multi-depth snowfall with varying particle sizes and aerodynamic settling speeds.
• Autumn Leaves: 3D physical fluttering trajectories, gentle rotational tumbling, and rich golden foliage colors.
• Gentle Smoke: Subtle curling vapor streams suited for meditation, candlelit ambiance, and coffeehouse aesthetics.
• Golden Dust Particles: Floating luminous bokeh specks providing a cinematic fantasy atmosphere.

Vietnamese - English Phonetic Normalizer

• Automatically normalize English loanwords, brand names, and tech acronyms (e.g. AI to "ây ai", ChatGPT to "chát giê pi ti", YouTube to "diu túp", TikTok to "tóp tóp", Facebook to "phây búc") into natural Vietnamese phonetics for TTS engines.
• Recognize and standardize informal Vietnamese messaging shortcuts (ko to "không", dc to "được", hnay to "hôm nay", ng to "người").
• Tag Protection System: Safely shield character speaking tags [Nam], [Nữ] and non-verbal emotion tags [laughter], [sigh], [SFX: ...] from unintended dictionary substitutions.
• Support Exact Match, Whole Word Boundary Match, and Regular Expression matching modes.

ElevenLabs Voice Engine

• Extended ElevenLabs API integration for studio-grade AI voice generation across 29 global languages.
• Automatic synchronization of custom cloned voices and community voice libraries directly into the application.
• Granular control over Stability, Clarity/Similarity, and Style Exaggeration sliders.

Smart Audio Loop & Mastering

• Automatic seamless crossfade looping for background audio tracks without audible clicks or abrupt breaks.
• Smart Ducking technology that dynamically attenuates background music when narration dialogue is present and restores volume during pauses.
• Professional audio mastering suite: Noise Gate, multi-band Equalizer filtering, and broadcast-standard LUFS loudness normalization (-14 dB for YouTube / -16 dB for TikTok).

Non-profit organizations, charities, and community initiatives supporting elderly people, children, people with disabilities, and other vulnerable groups are always welcome to reach out to us.

If your organization uses Perfex CRM and requires any of our modules, plugins, or related products, we will gladly provide assistance with installation, customization, training, and long-term operation.

As part of our commitment to supporting meaningful social impact projects, we offer lifetime free support for all products used by eligible non-profit organizations. Our goal is to help you streamline operations, reduce technology costs, and focus more resources on serving your community.

For inquiries and support, please contact us at polyxgo@gmail.com. We will do our best to assist your organization as quickly as possible.