About Scribeus

Part of the RuntimeHub free tools suite

Scribeus is a free, browser-based text-to-speech tool that converts written text into natural-sounding audio — entirely on your device. No data is sent to any server. No account is required. No usage limits apply.

How it works

When you first visit Scribeus, it downloads the Kokoro-82M AI model directly to your browser (~155 MB). This happens once; on every subsequent visit the model loads instantly from your browser's local cache. After that, all speech synthesis happens locally — the model runs inside a Web Worker, isolated from the page, processing your text without any network activity.

Long texts are automatically split at sentence boundaries into chunks of up to 450 characters. Each chunk is synthesized sequentially and stitched into a single WAV file. You can start listening as soon as the first chunk finishes — you do not need to wait for the full text to complete.

The model: Kokoro-82M

Scribeus is powered by Kokoro-82M v1.0, an open-source neural TTS model with 82 million parameters developed by hexgrad and distributed under the Apache 2.0 license. For browser inference we use the model_q4f16.onnx variant — a 155 MB quantized ONNX build that balances audio quality and inference speed well on consumer hardware.

ONNX inference is handled by ONNX Runtime Web (ORT), running in a Web Worker using the WebAssembly execution provider with SIMD acceleration when available.

Voices

Four American English voices are available:

• Bella — Female, warm and clear
• Jessica — Female, crisp and articulate
• Liam — Male, natural and conversational
• Adam — Male, deep and steady

Voice style vectors (speaker embeddings) are loaded on demand and cached locally alongside the model.

Privacy architecture

Scribeus is built on a strict zero-upload principle shared across all RuntimeHub tools. The synthesizer runs in an isolated Web Worker; the page never has access to the raw model weights. Your text and the resulting audio exist only in memory and are discarded when you close the tab or generate new audio. Nothing is stored on any server.

Part of RuntimeHub

Scribeus is one tool in the RuntimeHub suite — a growing collection of free, privacy-first, browser-based utilities. Other tools include Lexiconic (audio transcription), PixelBoost (AI image upscaling), and ZeroBG (background removal).

All tools are free, ad-supported, and built with vanilla JavaScript — no frameworks, no trackers, no data collection.