Voice Packs 📦

A starter set of 13 curated blend recipes from the lab’s 55 generated combinations — ASMR whispers, audiobook narrators, anime characters, professional anchors, and more. Open any pack in the designer, tweak it on the map, and make it yours.

🇺🇸 Warm Flame

stable

American English · American English

Blend: af_heart(60%) + af_bella(40%)

🎛️ Open in Designer

🇬🇧 Uk Elegance

stable

British English · British English

Blend: bf_emma(60%) + bm_fable(40%)

🎛️ Open in Designer

🇺🇸 Audiobook

blend

American English · American English

Blend: af_heart(40%) + am_michael(30%) + bf_emma(30%)

🎛️ Open in Designer

🇺🇸 Pro Authority

blend

American English · American English

Blend: am_michael(50%) + am_fenrir(30%) + am_onyx(20%)

🎛️ Open in Designer

🇯🇵 Jp Cute Anime

blend

Japanese · Japanese

Blend: jf_nezumi(40%) + jf_gongitsune(30%) + jf_tebukuro(30%)

🎛️ Open in Designer

🇨🇳 Zh Deep Male

blend

Mandarin Chinese · Mandarin Chinese

Blend: zm_yunyang(50%) + zm_yunjian(30%) + zm_yunxia(20%)

🎛️ Open in Designer

🇨🇳 Zh Professional M

blend

Mandarin Chinese · Mandarin Chinese

Blend: zm_yunjian(45%) + zm_yunxi(35%) + zm_yunyang(20%)

🎛️ Open in Designer

🇨🇳 Zh Sweet Girl

blend

Mandarin Chinese · Mandarin Chinese

Blend: zf_xiaoyi(45%) + zf_xiaobei(35%) + zf_xiaoxiao(20%)

🎛️ Open in Designer

🇨🇳 Zh Warm Lady

blend

Mandarin Chinese · Mandarin Chinese

Blend: zf_xiaoxiao(50%) + zf_xiaoni(30%) + zf_xiaoyi(20%)

🎛️ Open in Designer

🇺🇸 Asmr Whisper

experimental

American English · American English

Blend: af_river(40%) + af_sarah(35%) + jf_gongitsune(25%)

🎛️ Open in Designer

🇺🇸 Eastwest F

experimental

American English · American English

Blend: af_heart(50%) + zf_xiaoxiao(30%) + jf_alpha(20%)

🎛️ Open in Designer

🇫🇷 Euro Delight

experimental

French · French

Blend: ff_siwis(35%) + if_sara(35%) + ef_dora(30%)

🎛️ Open in Designer

🇯🇵 Sakura Dream

experimental

Japanese · Japanese

Blend: jf_alpha(40%) + af_heart(40%) + jf_gongitsune(20%)

🎛️ Open in Designer

Why only blends?

Packs reproduce exactly through the designer’s weighted-average synthesis — every pack is a convex combination of real base voices, so what you hear matches the recipe. Vector-math, extrapolation, recombine, and PCA-sampled recipes are less predictable; they stay in the experiment lab with their full recipe metadata.

Frequently asked questions

What are voice packs?

Voice packs are curated blend recipes: weighted combinations of real Kokoro voice embeddings, generated offline from the lab’s 55 experimental recipes. Only convex blends are promoted here because they reproduce exactly in the designer.

How do I use a pack?

Click “Open in Designer” and the pack loads into the Voice Maker studio. Press Preview (the model downloads once, ~305MB) to hear it, tweak it on the map, save your version, or download the .bin style file.

Why are some packs labeled experimental?

Packs that mix distinct language families (e.g. English + Japanese + Mandarin) are labeled experimental: they can sound less natural. Packs blending voices from one language family are labeled stable or blend.

Are packs free and do they require an account?

Everything is free with no signup or API key. Recipes are just numbers; the designer rebuilds each pack from the real base-voice style files in your browser.

Can I modify a pack?

Yes — after loading a pack you can keep dragging the map or adjusting traits, then save your version to My Voices or share it as a link.