AI Voice Studio is a professional-grade audio production environment designed for high-fidelity documentary narration. It bridges the gap between mechanical Text-to-Speech and genuine human-like performance by connecting your production pipeline directly to frontier AI models.
Download on the Releases Page Download Releases Or TRY THE APP HOSTED ON GOOGLE CLOUD
- Multi-Engine Synthesis: Seamlessly switch between Gemini 2.5/3.0, ElevenLabs Multilingual v2, and Resemble AI.
- Universal Translation Layer: Utilizes Gemini 3 Pro to translate your script while preserving SSML tags and emotional cues before synthesis.
- Vocal Identity Cloning: Create instruction-based voice profiles. Analyze your own voice samples to generate "Identity instruction" sets for the Gemini TTS engine.
- Long-Form Production: Automatic script chunking for narrations exceeding 10,000+ characters with gapless concatenation.
- Pronunciation Dictionary: Define custom rules for technical terms, names, or localized jargon.
- Store-Ready Architecture: Full support for Mac App Store (MAS) Sandboxing, Microsoft AppX, and Linux Flatpak distribution.
The Universal Translation Layer and multi-engine synthesis now support the following 41 languages and classical/indigenous dialects:
- Arabic (AR)
- Aymara
- Catalan
- Cherokee
- Chinese (ZH)
- Danish
- Dutch (NL)
- English (UK & US)
- Flemish
- French (FR)
- Georgian
- German (DE)
- Guarani
- Hindi (IN)
- Indonesian (ID)
- Italian (IT)
- Japanese (JP)
- Kannada
- Khmer
- Korean (KR)
- Lao
- Latin
- Latvian
- Maltese
- Maya (Yucatec)
- Nahuatl
- Navajo
- Nepali
- Norwegian
- Polish (PL)
- Portuguese (BR)
- Quechua
- Russian (RU)
- Sinhala
- Spanish (ES)
- Swedish (SE)
- Tamil
- Thai (TH)
- Turkish (TR)
- Vietnamese (VN)
.png)
.pnghttps://github.com/AjarnSpencer/AI-Voice-Studio/blob/main/assets/ai--studio%20(4).png)
.png)
.png)
The "Custom Voice" feature uses a Zero-Shot Analysis technique. When you provide a sample, the Gemini model decomposes the audio into a parametric text-based instruction set (Pitch, Tone, Pace, Prosody). This instruction is then fed as a system prompt to the TTS engine to modulate the base voice into your "Identity."
When narrating in a non-English language:
- Source Script is passed to Gemini 3 Pro.
- SSML-Aware Translation is performed, ensuring
<break>and[Whisper]tags remain in the correct semantic position. - Target Language Audio is synthesized by your selected engine (ElevenLabs, Gemini, or Resemble).
# 1. Install Production Dependencies
npm install
# 2. Compile Web Application
npm run build
# 3. Package for Distribution (Windows/Mac/Linux)
npm run electron:package
"Perform your narrative, don't just generate it."
Developed by Ajarn Spencer Littlewood