Kokoro: Free Voices (With Names)

Kokoro-82M is a free, local, Apache-licensed text-to-speech model with ~28 named preset voices — each graded by the pack itself (A- down to D-). No API. No meter. No bill.

In this video THE CAST CARRIES IT: Bear's ElevenLabs clone opens, then eight named Kokoro voices introduce themselves and each teach one lesson — what it is, how it plugs into the same interface, the honest grade range (including the F+ voice), the casting field, the full roster, the previz use case, the no-cloning tradeoff, and the D- wink.

This is the third voice engine in the toolkit, after ElevenLabs (metered) and Suno (flat). Who speaks is taste — and taste is yours.

───────────────────────────────── 🛠 TOOLKIT generate_audio_kokoro.py ships with the Brutalist toolkit: https://www.brutalist.art/

───────────────────────────────── 📼 THE SERIES — Brutalist: Claude for Video Production Episode 1 — What Is Brutalist? https://youtu.be/HsRc6F_vbs8 Episode 2 — Installs, .env & Credentials Episode 3 — When Cowork Can Help Claude Code Episode 4 — Posting to YouTube (with Captions) https://youtu.be/S7rmHr36C74 Episode 5 — Suno vs ElevenLabs Cost Test https://youtu.be/HsRc6F_vbs8 Episode 6 — Kokoro: Free Voices (With Names) ← you are here Episode 7 — She Walks in Beauty https://youtu.be/Z2noM0YHeVA

───────────────────────────────── @NikBearBrown · brutalist.art

▶ The full series playlist — Brutalist https://www.youtube.com/playlist?list=PLG9H-C6rp5RU

Transcript

A lightly formatted transcript of the video, with timestamps so you can jump to any section on YouTube.

[0:00] This is my voice, the 11 Labs clone that's narrated every video in this series. It's good. It's also metered. So, for the rest of this video, I'm handing the microphone to a cast that costs exactly nothing. They live on this laptop and they have names. >> I'm Bella. I'm a Kakoro voice and 82 million parameter open model Apache licensed running locally on the same machine that's rendering these graphics. No a PI, no meter, no bill. I'm what free sounds like. I'm Sarah. Here's why we plug straight in. A voice engine in this toolkit is just something that fills one MP3 per beat and measures it.

[0:39] Same files, same durations, same beat sheet. Downstream. Nothing knows I'm not the expensive one. >> I'm Adam. On the official report card, I'm graded F+. The pack grades its own voices and it doesn't flatter us. They let me say this beat anyway. That's the honest record. Free has a range and you should hear the bottom of it. >> I'm Michael. The command is one line and the casting is one field. This very beat says voice. A M underscore Michael. Change the field, change the reader. That's the whole workflow. >> I'm Emma, one of the British voices. There are about 28 of us, American and British, graded from A to F by the pack

[1:17] Itself. You've just heard an A minus, a C plus, and an F plus. Casting is taste, and now taste is free. I'm George. The strongest use isn't the final cut. It's the draft. Free narration means real timings, real conforms, a watchable preview before a single paid credit is spent. Gray box for the ears. >> I'm Puck. Now the catch. Honestly, none of us is you. Kakoro doesn't clone. When the video carries your name and your channel, the voice should be yours, metered or flat. When it doesn't, we work free. >> And I'm Santa, graded Dminus. Ho ho. You get what you pay for except you didn't

[1:54] Pay. >> So the roster is three engines now. Metered the clone steadiest priced by the character. Flat [snorts] your voice through sunno one generation per video and free us. The machine will voice every beat either way. Who speaks is taste and taste is yours. That's the free voice option. Subscribe to follow the series and find the whole toolkit at www.brutalist.art. You bring the taste, the machine brings the hands. are now a cast.