Workflows/Generate speech locally with Kokoro-82M

Audio / hexgrad

Generate speech locally with Kokoro-82M

Write text, generate Kokoro-82M speech, edit and export a WAV in a local visual editor rather than running a custom Python script for every line.

EstimatedSource checked 10/14/2026Version: Reviewed easiest local route · 2026-10-14

LOCAL

Freshness 8 out of 100. Repository age: 658 days old, Dec 26, 2024. Measures age, not quality.

Popularity 95 out of 100. 10,468,205 downloads in 30 days; 7,256 lifetime likes. Checked Oct 15, 2026. Measures adoption, not quality.

What do these scores mean?

Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.

Freshness · release age

A new release starts at 100. The score halves every 180 days: 100 × 2^(−age in days / 180). When the release date is unavailable, a repository date is a fallback—not proof of when the model was released. Missing dates stay unscored.

Check the date source

Popularity · adoption

Downloads over the past 30 days carry 80% of the score; lifetime likes carry 20%. Both use a logarithmic scale, capped at 10 million downloads and 100,000 likes. If likes are unavailable, downloads carry all the weight. Missing download statistics stay unscored.

round(80 × min(1, log10(1 + downloads) / 7) + 20 × min(1, log10(1 + likes) / 5))

Check the statistics source

Scores as of Oct 15, 2026. Statistics checked Oct 15, 2026. Statistics are marked for refresh after 72 hours.Full scoring method

Before you begin

Required VRAM: 4 GB

Hardware

One short English sample with the publisher af_heart voice. Memory is estimated for the complete editor and model, not a speed benchmark.

Machine	Kokoro-82M · af_heart · KokoroGUI
NVIDIA GPU · 2–4 GB VRAM + 8 GB RAM	2 GB starting estimate; 4 GB offers headroom for the short sample.
AMD or Intel GPU PC · 8 GB RAM	Use the editor’s CPU route unless a compatible backend is independently verified.
Apple Silicon Mac · 8–16 GB unified memory	Local editor route; use CPU fallback if MPS is unavailable.
CPU-only PC · 4–8 GB RAM	Short speech generation works without a dedicated GPU; allow 8 GB for margin.

The 327 MB model is only one component. Python, pronunciation data, audio buffers and the OS add memory.

Tradeoffs

This guide uses the publisher af_heart voice; it does not clone a supplied recording.
The initial text and voice are American English; other languages require matching language and voice settings.
The script retains generated chunks before writing the combined file, so long passages need more host RAM.

LOCAL SPEECH GENERATIONSOURCE-LINKED SETUP

Generate speech locally with Kokoro-82M

After a one-time KokoroGUI install, select Kokoro and the stock af_heart voice, type two sentences, click Generate, listen on the timeline and export one WAV. The interface handles model download and later edits.

Choose a package Use the built-in workspace Check the result

BEFORE YOU BEGIN

The documented package

This setup uses the package documented for this task. Review its source and supported platforms before starting.

Save your machine in My Hardware to get an automatic starting choice. You can always choose any package yourself.

CURRENT SETUP

Kokoro-82M · af_heart · KokoroGUI · community-modified weights

Memory figures are planning estimates for two short sentences and do not establish a minimum. A working Kokoro/eSpeak install and first model download are required.

Freshness 8 out of 100. Repository age: 658 days old, Dec 26, 2024. Measures age, not quality.

Popularity 95 out of 100. 10,468,205 downloads in 30 days; 7,256 lifetime likes. Checked Oct 15, 2026. Measures adoption, not quality.

What do these scores mean?

Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.

Freshness · release age

Check the date source

Popularity · adoption

round(80 × min(1, log10(1 + downloads) / 7) + 20 × min(1, log10(1 + likes) / 5))

Check the statistics source

Scores as of Oct 15, 2026. Statistics checked Oct 15, 2026. Statistics are marked for refresh after 72 hours.Full scoring method

View weight source

THE WORKFLOW

Set up Kokoro-82M · af_heart · KokoroGUI on your machine

Choose where you want to run it. Use the built-in workspace; no workflow import is needed.

CHOOSE YOUR SETUPWindows

Install the local Kokoro editor

Install Python 3.11 and eSpeak NG for Windows. In PowerShell, run the linked project’s short one-time setup, then double-click run.bat. Speech generation and editing happen in the desktop window. This community editor uses publisher Kokoro-82M weights; its current release is a beta.

KokoroGUI installation and supported platforms eSpeak NG downloads Python downloads

PowerShell

git clone --branch 4.0.0-beta.2 --depth 1 https://github.com/CoffeeMethod/KokoroGUI.git
cd KokoroGUI
py -3.11 -m venv .venv
.\.venv\Scripts\python.exe -m pip install -r requirements.txt

Open a project and select Kokoro

Double-click run.bat to open KokoroGUI, then create a new project. Select the Kokoro engine and American English af_heart stock voice. If the engine is missing, install its dependency according to the project setup and restart the editor. The publisher model downloads on first use; confirm the download is complete and the voice is ready before generating.

KokoroGUI model setup and usage Exact Kokoro-82M weights and voice

Generate a short, reviewable sample

Type two short English sentences, assign the af_heart voice, and press Generate. Wait for the speech clip on the timeline, press Space to play it, and listen for both sentences and names. Edit words or pronunciation in the visual editor and regenerate only the changed clip.

KokoroGUI generation workflow

Export and replay the audio

Choose File → Export and save one WAV. Open the exported WAV in your audio player and verify both sentences are audible. File → Save stores the editable project separately; exporting is required to get a standalone sound file.

KokoroGUI export instructions

SUCCESS CHECK · Kokoro-82M · af_heart · KokoroGUI

A complete WAV from the local editor

File → Export produces a WAV that plays both supplied sentences in the chosen Kokoro voice. The project remains editable so a changed sentence can be regenerated and exported again.

KokoroGUI is a community-maintained beta and needs Python/eSpeak for installation. Its visual editing and export path is simpler for repeat use than the older custom-script workflow; the underlying Kokoro-82M weights remain the publisher’s.

Back to setup steps

When it doesn’t go to plan

The Kokoro engine or voice does not appear.

Install the Kokoro dependency and eSpeak NG using the editor’s platform instructions, restart it, choose Kokoro rather than Audio8, and wait for the first model download.

The first generation fails or is silent.

Try a two-sentence English sample with af_heart, confirm the model cache finished and play the generated timeline clip before exporting.

The editor plays speech but no audio file exists.

Use File → Export to create the WAV. File → Save stores an editable project, not the final audio file.

REFERENCE LIBRARY

Sources and files

Original instructions, model files, and compatibility notes behind this setup.

11 SOURCES

01Kokoro publisher setup and stock-voice APIgithub.com 02Exact Kokoro-82M v1.0 weightshuggingface.co 03Publisher stock voice af_hearthuggingface.co 04Kokoro 0.9.4 packagepypi.org 05KPipeline device and cache implementationgithub.com 06Misaki English frontendgithub.com 07eSpeak NG releasesgithub.com 08spaCy English modelspacy.io 09PyTorch platform wheelspytorch.org 10uv installationdocs.astral.sh 11KokoroGUI local editor, install and exportgithub.com

The model behind this workflow

Kokoro-82M

Keep exploring

Text

SETUP GUIDE

Qwen3.5 · 9B private chat setup

Source-backed package and platform-route plan for private local chat with Qwen3.5 · 9B. Preserves the official Ollama 9B package, a smaller 4B family fallback, and a clearly labeled community uncensored GGUF option.

Freshness 41 out of 100. Release age: 230 days old, Feb 27, 2026. Measures age, not quality.

Build a private coding assistant with Qwen3.8 27B

Run Qwen3.8 27B with Ollama, connect it to Cline in VS Code, and make one reviewable code change in a tiny local project.

Freshness 76 out of 100. Repository age: 71 days old, Aug 5, 2026. Measures age, not quality.

Arena Open Source Web coding: rank 10 of 55. Ranked model: qwen3.8-27b. Original model ranking, not a benchmark of this package. Export dated Oct 13, 2026. Source checked Oct 15, 2026. Human preference ranking, not hardware suitability.+3

SOFTWAREOllama + Visual Studio Code + Cline + Python

Create images locally with FLUX.2 Klein

Generate images from text prompts in ComfyUI with the fast distilled or more adjustable base 4B model. Save your results and reuse the visual workflow.

Freshness 35 out of 100. Release age: 273 days old, Jan 15, 2026. Measures age, not quality.

SOFTWAREComfyUI

LOCALRENTED GPU

Verified source