Whispering Tiger
Local AI transcription and translation software for VRChat and streaming overlays. Runs offline after model downloads, with an MIT license.
Whispering Tiger turns audio on your computer into live transcripts and translations for VRChat and streaming overlays. It's free, open source software under the MIT license, aimed at people who want captions or translated conversations while gaming or broadcasting. Processing stays on your machine, and it works offline once you've downloaded the models.
Its connection to other apps is a central feature. It sends text to VRChat through OSC or to browser overlays through WebSockets, with support for streaming applications such as OBS, vMix and XSplit. It can listen to microphone input and computer audio, so its use extends beyond transcribing your own speech.
Speech recognition options include OpenAI's Whisper, Meta's Seamless M4T and NVIDIA's NeMo Canary. For text translation, it supports NLLB-200 and M2M-100, alongside automatic language identification. CUDA GPU acceleration is available.
It also reads text in game images. EasyOCR handles text recognition, while GOT-OCR 2.0 and Microsoft's Phi-4 Multimodal LLM support handwriting. That gives players a way to translate visible text as well as spoken dialogue.
Text-to-speech can read transcripts and translations aloud. Supported engines include Kokoro TTS and Silero; F5/E2-TTS and Zonos add voice cloning and streamed playback. An RVC plugin can change voices from live audio, audio files or generated speech. A separate native interface manages settings and application updates.
