Keep their voice close — even when they can’t be.
Clone a voice you’re allowed to keep — parents, friends, mentors — for stories, reminders, and everyday moments. Not a replacement for people. A way to carry their warmth.
I can’t always have them here. I can still hear them.
Parents
A lullaby, a prayer, a “drive safe.”
Friends
Pep talks, inside jokes, the voice that calms you down.
You
Your own voice, for the days you’re tired of typing.
Only use voices you have permission to keep.
Mac desktop · hosted studio
Yaazh
Voice, Generate, Studio, Live, Instagram — on this device.
The same studio as the app: create a voice, generate speech, mix stories, and keep captures. Only use voices you have permission to keep.
Create a voice
Clone a voice you have permission to keep from a short recording, or start with built-in Adam, Bella, Heart, and Tamil Rasa. Family voices ask for consent first.
Generate speech
Type, pick a voice, hear them. New sessions default to Adam · English · Kokoro. Clones use your samples — empty clones cannot generate.
Stories Studio
Multi-voice narratives on a timeline. Arrange tracks, generate lines, and mix conversations between people you have permission to keep.
Live & Dictate
Speak into Live or Dictate. Whisper transcribes on this device. Captures keep the clip and the transcript together.
Effects
Yaazh packs — Phone call, Warm hug, Radio host, Storytime, Temple hall, Night whisper — plus classics. Preview dry/wet before you save.
Share a lullaby or a prayer as a Story or Reel. Copy captions in Tamil or English, then open Instagram. Yaazh never posts for you.
On your machine
Mac app (Coming soon) downloads models after install — never inside the DMG. Or try the hosted studio. No account. No cloud backup of your voices.
Any clip becomes a voice — if you have permission to keep it.
Three ways to get a sample in. Upload a clip, record from your microphone, or capture audio playing on your system. Yaazh clones the voice from as little as 3 seconds of audio. Only use voices you have permission to keep.
Click to record from your microphone.
Maximum duration: 30 seconds.
Dictate, Live, and Captures.
Speak in the studio. Transcripts stay on this device. Captures land with the original clip so you can replay them later.
Dictate and Live
Speak in the studio. Whisper transcribes on this device. Live is a conversation loop; Dictate is for a single take.
Captures keep the clip
Each capture stores audio plus transcript. Replay, copy, or use a clip when you create a voice — only with permission.
Agents Preview
Assign a voice, then Preview. Agents speak through the same pill as Live — never a silent no-op. MCP tool names stay voicebox.speak.
Share a voice from home. Yaazh never posts for you.
Export a Story or Reel — a lullaby, a prayer, a “drive safe.” Copy captions in Tamil or English, then open Instagram on your phone. Only share voices you have permission to keep.
யாழ்
Amma’s lullaby
A voice from home
Instagram Story
Tamil · English captions
Open Instagram
Agents speak in a voice you assigned.
Preview waits for the generation, then plays it. One tool call — voicebox.speak — and Cursor or Claude Code can talk in a voice from your studio. Assign a voice first, or Preview stays disabled.
{
"mcpServers": {
"yaazh": {
"url": "https://app.yaazh.nexageapps.com/mcp"
}
}
}// In any MCP-aware agent:
await voicebox.speak({
text: "Deploy complete.",
profile: "Morgan",
})POST /speak for anything that doesn’t speak MCP — ACP, A2A, shell scripts, or custom harnesses.Per-agent voice
Bind each agent to a voice in Settings. Preview uses that voice — or the MCP default — never a silent Idle chip.
Always visible
Every agent-initiated speech surfaces the pill. No silent background TTS — you always see what’s coming out of your machine.
Open protocols
Agents Preview is in the studio today. MCP tool names stay voicebox.speak; the server advertises as yaazh.
Voices with a personality.
Give any voice profile a free-form personality. Then Rewrite your text in their voice, or let them Compose a fresh line of their own — your cloned voice, in full character.
“Warm Tamil mother. Speaks as if you are still at the door. Mixes English with a little Tamil. Reminds you to eat, drive safe, and call when you reach.”
Rewrite
Restate your text in their voice while preserving every idea. Same content, their delivery — a reminder, a prayer, a line in a story.
Compose
No input needed — Compose improvises a fresh line in their voice. Roll again for another take. A lullaby, a pep talk, a “drive safe.”
Your local voice API
Every engine you download becomes a REST endpoint on your machine. Build apps, games, and voice tools with full programmatic control — no API keys, no rate limits, no per-character fees.
https://app.yaazh.nexageapps.com/generateGenerate speech/generate/{id}/cancelCancel a generation/profilesList voice profiles/profilesCreate a new profile/models/statusModel catalog & state/historyPast generations/healthServer healthcurl -X POST https://app.yaazh.nexageapps.com/generate \
-H "Content-Type: application/json" \
-d '{
"text": "Welcome to the game, player one.",
"profile_id": "b3f1c2d4-5e6f-4a7b-8c9d-0e1f2a3b4c5d",
"engine": "qwen_custom_voice",
"instruct": "warm, slow, cinematic"
}' \
--output line.wavGames
Generate NPC dialogue on the fly, localize characters into new languages, or ship expressive voice lines without a studio.
Apps & agents
Give your app or AI agent a voice. Real-time narration, accessibility readouts, voice replies — all running on the user's machine.
Scripts & tools
Batch-generate audiobook chapters, automate podcast intros, or wire Yaazh into your Stream Deck. It's just a localhost URL.
Models you actually use
Download them in Models after install. The Mac app is small; weights live under Application Support — never inside the DMG.
Built-in speech
Ready after you download the weights. Not packed inside the Mac DMG.
Kokoro
by hexgradAdam, Bella, and Heart. New Generate sessions default to Adam · English · Kokoro so a first line is not “voice missing.”
Piper Rasa
by tinisoftBuilt-in Tamil default — a real female speaker, not pitch-shifted MMS. Tamil clones stay on Qwen with your sample.
Cloning
Your recording. Empty clones cannot generate.
Chatterbox Turbo
by Resemble AIDefault for English clones. Presence and pace sliders. CPU Spaces skip the unused 1.1GB s3gen file.
Qwen3-TTS
by AlibabaTamil clones keep your speaker. Hosted studio uses 0.6B, not 1.7B. Delivery instructions when you need tone and pace.
Listen & clean up
Speech → text in Live, Dictate, and Captures.
Whisper
by OpenAIOn-device transcription. Pick a size that fits this machine. Captures keep the original clip next to the transcript.
Qwen3 instruct
by AlibabaOptional transcript cleanup and persona rewrite. Caps at a few seconds so Generate is not starved.
Keep it on your Mac
Free to try. Local to keep. $0.99 lifetime for the first 1M. No account · No ads · No server · Privacy first. Download stays Coming soon until a notarized DMG exists.
Join Mac beta