One owner, one model
A focused model represents one authorized voice. It does not carry a catalog of voices the owner will never use.
VoiceYog™ · Personal TTS · Apple Silicon
Train your voice once. Use it for years.
We start with one authorized voice sample and turn it into a lightweight, single-voice model that runs locally on supported Apple Silicon Macs. Generate speech without a cloud voice API—and reuse the same voice in presentations, lessons, narration, and future applications.
Your voice model is built for you. Training uses our GPU lab once; everyday speech generation runs on the Arm CPU already inside your Mac.
Voice ownership for working creators
Stop recording the same voice work over and over.
If you produce lessons, training videos, presentations, tutorials, or audiobooks every month, your voice is part of your production workflow. VoiceYog™ is designed to turn one authorized sample into a reusable personal voice model—so you can update and create narration locally without repeatedly recording every line or depending on a cloud voice API.
Voice preservation is an important second use case.
A person's voice can change because of illness, treatment, injury, stroke, or aging. Some people want to preserve how they sound today and retain a voice they can use for future content. VoiceYog™ is a personal content-creation technology—not a medical device or treatment.
Local for everyday use
After installation, inference runs without CUDA, PyTorch, a cloud voice API, or a network connection.
One voice, many uses
Use the model for presentation videos today and connect the same local voice engine to other content workflows later.
The VoiceYog™ model lifecycle
Sample → Train once → Install → Reuse
The expensive model-building work happens once on DLYog Lab GPU infrastructure. Repeated speech generation moves to the user’s Apple Silicon CPU.
01 · SAMPLE
Capture the voice
One authorized voice sample is enough to begin building the focused training corpus.
02 · TRAIN
Build the specialist
Our GPU lab builds and validates a compact model focused on that single voice.
03 · INSTALL
Move it to the Mac
The exported model and Arm-aware runtime are installed on supported Apple Silicon.
04 · REUSE
Generate locally
Create speech offline through VoiceYog™ Video Studio or other applications connected to the local engine.
Watch the demo
See VoiceYog™ speak for itself.
Watch the model, local Arm inference workflow, and the technical idea behind training once and generating speech repeatedly on the user's own hardware.
Open the demo on YouTubeFirst application · VoiceYog™ Video Studio
Turn PowerPoint speaker notes into narrated video.
VoiceYog™ Video Studio puts the personal model into an end-to-end Mac workflow. Bring the deck you already made, control every slide and line, and export a finished video without recording every take from scratch.
Start with the existing deckExtract slides and speaker notes instead of rebuilding content in a new editor.
Review every narrationEdit notes, listen slide by slide, regenerate, or replace audio with a real recording.
Change without starting overInsert supporting media and rebuild only the sections that changed.
More from DLYog Lab
Existing products and technical work.
Personal voice ownership is our current product direction. These existing projects remain available without competing with that primary message.
VoiceYog™ reference model
The released 68.5 MB reference model demonstrates the single-voice architecture, Arm-aware runtime, offline inference, and reproducible benchmarks behind the personal-model direction.
Technical note: Personal voice model · Arm-optimized TTS · Private, local inference · Developed and benchmarked in DLYog Lab's in-house AI lab.
AI4Echolalia™
A caregiver-support product for use between speech-language therapy sessions. Available to existing and invited users and for licensing or strategic partnership.
Project recognition: TED AI Hackathon, 2nd Place · 2024. This recognition applies to AI4Echolalia™, not every DLYog Lab product.
Open AI4EcholaliaSynYog AI
A synthetic-data and model-development platform with assisted schema generation, multimodal processing, training, evaluation, and API inference.
Product note: A private-preview synthetic-data platform, separate from the VoiceYog™ personal voice product.
Explore SynYog AIEarly personal voice pilot
What would you create with a voice that stays on your Mac?
We are looking for Apple Silicon users who regularly create lessons, presentations, tutorials, audiobooks, or other narrated content—and want a personal voice they can reuse without depending on a cloud voice API.
AI for Good remains our compass: useful AI should be open, accessible, and under human control.