Paradee-8M v1.2

Paradee-8M v1.2

A tiny English text-to-speech model with four voices that runs on a CPU from one 8.5 MB ONNX file. For developers who want speech in an app without a GPU.

Made by sahilmahendrakar on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Paradee-8M v1.2 turns English text into speech. The model holds 7.88M parameters and learned from Kokoro-82M, a larger voice model. Version 1.2 speaks in four voices, two American and two British. The int8 model is a single 8.5 MB ONNX file that holds all four voices.

Who it suits: Developers who want spoken audio from text on a plain CPU, in an app or a script, without a GPU or a large download.

Industry | Who uses it | How they use it | Example
Software | App developers | Add a voice to a tool that runs on a laptop | An app that reads its messages aloud
Accessibility | Tool builders | Read screen text aloud on modest hardware | A small reader for notes
Media | Hobby creators | Make short English voice clips offline | A clip saved as hello.wav
Education | Teachers | Read short English passages in a British or American voice | A spoken example sentence

You can skip it if you need languages other than English, or voice quality at the level of its larger teacher model.

Where to use it:

• Python scripts, through the paradee package
• The command line, with python -m paradee
• Browsers and Node, where the tokenizer file loads in transformers.js and kokoro-js
• Windows, macOS or Linux, on one CPU thread

Use case | Who | What they get back
Offline narration | Creators | A WAV file read in the voice you pick
App speech | Developers | Audio from a small local model, about 62 times faster than real time on one thread
Voice choice | Developers | American or British, male or female

How to set it up:

• Python and pip on your computer
• The paradee Python package, installed from the project's GitHub repository with the command on the card
• The soundfile package for the Python example

1. Install the paradee package using the install command on the model card.
2. Optionally download this repository with the hf command.
3. Run the command line example or the Python example.

pip install -U huggingface_hub
hf download sahilmahendrakar/Paradee-8M-v1.2

How to use it: Run the module with your text, a voice name and an output file, or call it from Python. The card's Python snippet loads version 1.2 by default.

Tool | What it does
python -m paradee | Speaks a line of text and saves it to a WAV file
Paradee class | Loads the model in Python and returns audio for a text and voice
version argument | Loads an earlier release, such as 1.1, instead of 1.2

python -m paradee "Paradee is a small voice that runs anywhere." --voice bf_isabella -o hello.wav

from paradee import Paradee, SAMPLE_RATE
import soundfile as sf

tts = Paradee()   # v1.2; Paradee(version="1.1") loads the previous release
sf.write("hello.wav", tts("Paradee is a small voice that runs anywhere.", voice="bm_fable"), SAMPLE_RATE)

Example: You run the command with a short sentence and the voice bf_isabella. It writes hello.wav with that sentence spoken in a British female voice.

For newcomers and experts: If you are new, run the command line example and play hello.wav. Then change the voice name to hear the other three.

If you know the field, version 1.2 gives the text side and the decoder side a lookup of per-voice vectors, and the voice you name selects the entry. The British voices use British phonemes through misaki. The card reports UTMOS scores close to its Kokoro teacher on four voices, for example 4.40 against 4.52 for af_heart.

Text goes in
Phonemes, American or British
Small model with a per-voice vector
WAV audio comes out

Title: File size in MB
Paradee int8: 8.5 MB
Paradee fp32: 35 MB
Kokoro: 325 MB

Good to know:

• License: Apache 2.0
• Cost to run: none named, and the card says it needs no GPU
• Data it reaches: the files you download, and the text and audio files you give and write
• Voices: af_heart, bf_isabella, am_puck and bm_fable
• Not for: languages other than English, since the card describes an English model

On Hugging Face, 13 November 2026

• Task: text to speech
• Library: onnx
• Downloads in the last 30 days: 11,405
• Likes: 28
• License: apache-2.0
• Languages: en
• Based on: hexgrad/Kokoro-82M
• Tags: onnx, paradee, text-to-speech, distillation, kokoro

Links

• Model: https://huggingface.co/sahilmahendrakar/Paradee-8M-v1.2
• Files: https://huggingface.co/sahilmahendrakar/Paradee-8M-v1.2/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download sahilmahendrakar/Paradee-8M-v1.2

Then follow the usage in its model card. Nothing in this kit runs by itself.