Single voices are fine. But blended voices? That's where the magic happens.
You've probably used one or two of the 54 built-in Kokoro voices by now. They're solid. But here's what most people don't realize: you can combine any voice with any other voice to create something entirely new. You're not limited to 54 choices—you have hundreds of combinations at your fingertips.
The syntax is simple. Pair voices with a + sign, add weights in parentheses, and boom—you've got a custom voice that sounds like nothing else out there. The best part? You don't need to understand the theory. I'm giving you five recipes that actually work, tested and ready to go.
The Blending Syntax (Really Simple)
Before we jump into recipes, here's all you need to know:
Equal blend: af_sarah+af_bella — This mixes two voices 50/50.
Weighted blend: af_sarah(60)+af_bella(40) — First voice gets 60%, second gets 40%.
Multi-voice blend: am_michael(60)+bm_lewis(25)+af_heart(15) — Three or more voices, each with a weight. The percentages add up to 100.
That's it. Copy a recipe, paste it into Vois, and listen. If it doesn't feel quite right, try the next one. You can't break anything—you're just listening.
Recipe 1: The Warm Podcast Host
The blend: af_sarah(65)+af_sky(35)
What it does: Bright, energetic, but not overwhelming. Perfect if you're hosting a podcast and want listeners to actually trust you.
af_sarah is warm and approachable—she sounds like someone you'd grab coffee with. af_sky adds just enough sparkle and clarity so you cut through when there's background noise. Together? You sound like someone who actually has something interesting to say.
Use it for:
- Podcast hosting
- YouTube tutorials
- Friendly educational content
- Product launches
Speed tip: Run this at 1.04x speed. It sounds more natural for conversational content and keeps energy up without sounding rushed.
Recipe 2: The Authoritative Narrator
The blend: am_michael(60)+bm_lewis(40)
What it does: Serious, trustworthy, deep. This is for when you need listeners to believe what you're saying.
am_michael is an American male voice with genuine warmth, but bm_lewis adds that crisp British authority. Combine them and you get gravitas without coldness. It's the voice of someone who's done the research, knows what they're talking about, and genuinely cares if you understand it.
Use it for:
- Non-fiction audiobooks
- Business podcasts
- Documentaries
- Educational content
- Serious analysis
Speed tip: Slow this down to 0.94x speed. Authority lives in the pause. People respect slow, deliberate delivery.
Recipe 3: The Friendly Teacher
The blend: af_heart(55)+af_nova(45)
What it does: Warm but clear. Caring but professional. This is what good teaching sounds like.
af_heart is naturally warm—the kind of voice that makes you feel understood. af_nova brings brightness and clarity so technical terms don't disappear. The mix is perfect for anyone explaining something complicated to someone who might be intimidated. You sound like you actually want them to get it.
Use it for:
- Online courses
- Skill tutorials
- Explainer videos
- Language learning
- Fitness content
Speed tip: Keep this at 1.0x. Natural speed with warm voices sounds approachable and patient.
Recipe 4: The Dramatic Storyteller
The blend: af_heart(50)+bf_lily(50)
What it does: Elegant, nuanced, emotionally available. This is for stories with real feeling.
af_heart brings warmth and intimacy. bf_lily is British refinement and slight distance. When you blend them 50/50, you get something that can whisper confidentially one moment and command a room the next. It's perfect for fiction, especially if you want listeners to actually feel something.
Use it for:
- Fiction audiobooks
- Short story collections
- Narrative documentaries
- Guided meditations
- Emotional storytelling
Speed tip: Try 0.96x speed. Slower delivery with this voice blend creates space for emotion. Listeners need time to feel.
Recipe 5: The News Anchor
The blend: bf_emma(50)+bf_alice(50)
What it does: Professional, neutral, credible. The voice of someone reporting facts you can trust.
Both voices here are British English, both crisp and clear. But they have slightly different qualities—one's a touch brighter, one's a touch deeper—and blending them creates this interesting neutral middle ground. Not warm, not cold. Just completely professional and reliable.
Use it for:
- News and current events
- Business reports
- Financial updates
- Science and research summaries
- Official announcements
Speed tip: Keep this at 1.0x. News anchor energy is about clarity, not speed or slowness.
Building Your Own Recipes
Okay, you've got five recipes to try. But here's what happens after you use them for a bit: you start thinking about what you actually want. Maybe you like the warmth of Recipe 3 but need the authority of Recipe 2. Or you want Recipe 4 but a little bit more energetic.
This is where you start experimenting. And it's honestly not that hard.
Here's the pattern: Pick a dominant voice (the one that's 50%+ of the blend) that matches the feeling you want. Then pick a second voice that adds something the first one is missing.
Want warm but authoritative? af_heart(60)+am_michael(40) — warm with underlying authority.
Want bright but trustworthy? af_sky(55)+am_adam(45) — energetic without sacrificing credibility.
Want professional but approachable? bf_emma(55)+af_sarah(45) — precise British with American warmth.
The formula is: dominant voice (your base character) + complementary voice (what's missing) = something new.
The One Thing to Avoid
You can blend as many voices as you want, technically. But here's my honest recommendation: stick with two or three voices max.
More than three voices and you risk muddy sound. Everything blurs together. You lose character. It's like adding too many colors to a painting—instead of something beautiful, you get brown.
Two voices is the sweet spot for most content. Three voices if you really know what you're doing. Beyond that? You're probably overthinking it.
Test Before You Commit
Before you render a whole project with a new blend, test it:
- Pick a voice blend from above (or create your own)
- Generate a 10-20 second sample with actual text from your project
- Listen. Does it sound like what you wanted? Does it fit the content?
- If yes, go ahead and use it everywhere
- If no, try a different recipe
That's it. Five minutes max. You'll know instantly if it works.
Common Questions
Q: What if none of these recipes feel right?
Start with the one closest to what you want, then adjust. If Recipe 1 is too bright, try bumping the first voice down to 60% and the second voice up to 40%. Small changes matter. You don't need to be scientific about it—just listen and feel.
Q: Can I use the same blend for different projects?
Absolutely. Some of these blends (especially the narrator and teacher recipes) are genuinely versatile. You can use them across projects, save them by name, and build a personal library of trusted voices. Smart move.
Q: How do I know which voice is which in the syntax?
The first part before the parentheses is the voice ID (like af_sarah). The number in parentheses is the weight. Check the Vois voice library if you're not sure which voice is which—listen to samples before you blend.
Q: What if I blend voices from different languages?
You technically can, but the results are unpredictable. American + British English blends work beautifully. Mandarin + English? Probably not. Stick to voices from the same language family for predictable results.
Here's the thing: you came here with 54 voices to choose from. You're leaving with hundreds of combinations. Pick one of these five recipes and use it for your next project. Pay attention to how it feels. Then pick another one.
After a few projects, you'll develop instincts. You'll know what "warm + authoritative" actually sounds like before you ever generate audio. You'll understand which voices play well together. And that's when you start creating your own recipes.
That's the craft.