Yiperi Voice | Stylized Speech Engine

Yiperi Voice is a lightweight speech engine that converts text into charming, stylized gibberish speech in real-time. It provides a retro, cozy game aesthetic while solving the massive development bottleneck of voice acting and localization for large text systems.

Designed for cartoon, retro, comedy, RPG, creature, NPC, or UI dialogue. This is not a full dialogue manager and not a resources intensive realistic text-to-speech.

Problems This Solves:

1. Zero Voice Actor Budget: Give unique voices to dozens of NPCs using simple text strings.
2. No Multi-Language Audio Localization: Since the output is stylized gibberish, your dialogue audio fits any translated language automatically.
3. Completely Offline & Lightweight: Runs entirely local with zero external API dependencies, monthly subscription fees, or internet required.

Features:

✅ 12 built-in voice banks authored at home: Cartoon, retro, robotic, whisper, keyboard, mosquito, zombie, and metallic-style voices.

✅ Offline real-time playback: Plays through a normal Unity AudioSource.

✅ Text analysis into words, tokens, pauses, and stress markers.

✅ Word callbacks for syncing text reveal with speech.

✅ Token and viseme callbacks for custom animation or effects.

✅ Language definition files based on editable JSON.

✅ Supports adding custom voice banks by recording up to 68 phonetic tokens.

✅ Includes demo scenes for dialogue, minimal setup, and voice/language showcase.

Current language support: English-en, Spanish-es, Japanese-ja, Polish-pl, Thai-th, Arabic-ar, French-fr, Korean-ko, Portuguese-pt, Turkish-tr, Czech-cz, Hindi-hi, Dutch-nl, Russian-ru, Yiperi-yiperi, German-de, Italian-it, Norwegian-no, Swedish-sv, Chinese-zh

Language support means broad approximations designed for stylized cartoon speech, not exact linguistic pronunciation. For native writing systems where pronunciation cannot be reliably guessed, Yiperi supports [display<>say] syntax, so your UI can show native text while the engine reads a pronunciation-friendly version.

How It Works: The system uses "voice tokens", a simplified phonetic system that generates procedural audio clips based on the text strings you pass to it. Perfect for retro RPGs, cozy simulators, city builders, and single-player prototypes where fully voiced dialogue is unfeasible.

Included demos:

• Demo-1-Dialog: animated character dialogue showcase.
• Demo-2-minimal: minimal script setup.
• Demo-3-Voices: voice and language showcase.

Additional notes:

• Text Mesh Pro Essentials may be required to view UI text in the demo scenes.
• Python scripts are avaiable to help building custom voice banks.

DOCUMENTATION | FORUM | E-MAIL | SOCIAL | YOUTUBE