Coming soon

About This Software

Press a hotkey, speak, and the text lands at your cursor

ORI is a real-time voice dictation tool for Windows. Press the global hotkey and start talking, and recognition appears word by word right at your cursor — any text box, editor or chat window works, and so do games.

Highlights

• Live typing: words appear as you speak; when the engine revises a sentence it backspaces and retypes it, or you can switch to sentence-by-sentence output.

• Never drops the first word: the microphone is pre-warmed, plus about 400 ms of pre-roll buffering, so the opening is never lost.

• Noise-tolerant segmentation: energy-based voice activity detection cuts sentences, with automatic noise-floor calibration at the start of every recording.

• Undo a mistake: press Delete or the left mouse button to interrupt — everything not yet typed is discarded.

• Push to talk: hold the hotkey to speak, release to stop and type immediately.

• Auto stop: stops after 10 seconds of silence; a single recording is capped at 60 seconds (both adjustable).

• Game mode: fits games' custom-drawn input boxes via system-level clipboard paste.

Built-in offline engine

The Steam version ships with a local recognition engine, ready the moment it's installed: no account, no API keys, fully offline, no per-use cost.

Please note: the built-in offline engine currently supports Chinese speech only; for English speech input, switch to a cloud engine in the settings (bring your own key).

Higher-accuracy cloud engines (bring your own keys)

• Alibaba Cloud Bailian (Qwen speech recognition)

• Tencent Cloud real-time speech recognition

• iFlytek real-time transcription (LLM / standard editions)

• Volcengine Doubao streaming speech recognition

Switch from the right-click menu at any time, no restart needed. Each vendor offers free trial quota; cloud services are opened and billed by you directly with the vendor — this software never resells or bills them.

Text injection and compatibility

Two ways to type the result: paste via clipboard, or simulate keystrokes character by character. The clipboard is automatically protected and restored. Game mode forces clipboard injection to fit custom-drawn input boxes; when the target app runs as administrator, elevate ORI in one click, or always run it elevated.

AI correction and voice commands

• AI correction: fixes homophone typos, adds punctuation, removes filler words, with a customizable prompt; compatible with any OpenAI-compatible endpoint (DeepSeek, OpenAI, Moonshot, Zhipu, local Ollama, etc.).

• Voice commands: say "listen to me, select all" or "listen to me, undo" for local commands that cost no model calls; say "help me polish / translate / expand" to run the AI on the selected text; custom phrases insert with a single word.

End-to-end AI voice chat

Beyond dictation, you can talk directly to an AI: Doubao's end-to-end speech model, Alibaba Cloud Qwen-Audio, or a self-hosted Hermes Agent. Cut in at any time while the AI is speaking; the floating bar shows the conversation as subtitles.

Interface and system integration

• Always-on-top floating bar: four themes (light / dark / minimal / pixel), with adjustable font, position, size, opacity and auto-hide.

• History: recognition results can be recorded, with a separate window to search the full history.

• Automatically pauses background media playback or mutes the system while recording (optionally in game mode only), and restores it when you stop.

What the Steam version includes

• Built-in offline recognition engine and dependencies — out of the box, zero setup

• Steam Cloud saves: by default only a sanitized config (no keys) and history are synced; syncing keys is opt-in

• Stats: only aggregate numbers (recording duration, characters recognized, session count) — no text or audio

• Exclusive pixel-art theme

Privacy

With the built-in offline engine, audio and recognition results are processed entirely on your machine and never leave it. With cloud engines or AI correction, audio or text is sent directly to the provider you configured.

AI Generated Content Disclosure

The developers describe how their game uses AI Generated Content like this:

During use, the speech recognition and AI features connect to cloud services configured by you (Alibaba Cloud / Tencent Cloud / iFlytek / Volcano Engine speech recognition, or OpenAI-compatible LLM endpoints), and the resulting text is generated in real time by the selected service.