woheller69/whisperIMEplus

Android Input Method Editor (IME) based on RTranslators Whisper implementation

What it solves

Whisper+ is an Android input method editor (IME) that provides high-accuracy voice-to-text transcription and translation, allowing users to dictate text into any app without relying on cloud-based services.

How it works

The app leverages the Whisper engine (using ONNX models) to perform voice recognition entirely offline on the device. It can be used as a standalone application for transcription and translation to English, or integrated as a system-wide voice input service (RecognitionService) and an IME that can be triggered via other keyboards like HeliBoard.

Who it’s for

Android users who want a private, offline voice recognition and translation tool that can be integrated directly into their system's text input flow.

Highlights

  • Offline Processing: Voice recognition works entirely on-device for privacy and convenience.
  • System Integration: Functions as both an IME and a system-wide RecognitionService supporting RecognizerIntent.ACTION_RECOGNIZE_SPEECH.
  • Multilingual Support: Supports multiple languages with the ability to predefine two languages for quick switching and translation to English.
  • Flexible Deployment: Can be activated as a standalone app or integrated into other keyboards.

Related

  • Project
    openwong2kim/wmux

    wmux is a cross‑platform desktop app (Windows/macOS) that lets you run multiple CLI‑based AI agents (Claude Code, Codex, Gemini, etc.) in split panes, keep their sessions alive via a background daemon, and answer their interactive prompts from an iPhone companion. It adds Git worktree fan‑out, built‑in browser automation, a global “Fleet” view, and deep GitHub integration, making it an Agent Development Environment for AI‑coding agents.

  • Project
    Rvosy/Sakura

    Sakura Desktop Pet is an open‑source, cross‑platform framework for AI‑driven desktop companions. Users create character packs (art, personality, voice) and connect a vision‑language model service; the pet periodically screenshots the screen, lets the model decide whether to speak or act, and can be extended with plugins (e.g., browser automation).

  • Project
    ankur-anand/unisondb

    UnisonDB is an open‑source, log‑native real‑time database written in Go, aimed at edge‑AI and edge‑computing workloads. It stores data in a B+‑Tree (BoltDB or LMDB) while exposing the write‑ahead log (WALFS) as a streaming replication source. Replicas can pull the log via gRPC or object‑store (S3/MinIO/GCS/Azure), enabling sub‑second fan‑out to thousands of edge nodes. The engine supports key‑value, wide‑column, and large‑object models, atomic multi‑key transactions, and Zero‑MQ change notifications. Benchmarks show 60‑80 k writes / sec (1 KB) with low latency, making it suitable for reactive, distributed AI pipelines.

  • Project
    modular/mojo-gpu-puzzles

    An interactive, open‑source tutorial that teaches GPU programming in Mojo through hands‑on puzzles, complete with a `pixi`‑based environment, test harnesses, and community support.