Voxly

Voxly is a context-aware dictation app designed for people whose speech includes names, technical terms, niche vocabulary, multilingual phrases, or project-specific language that generic transcription tools often misunderstand. Instead of treating dictation as a one-size-fits-all transcript, Voxly lets users provide personal context and maintain a custom vocabulary so their spoken words can be corrected into polished, accurate writing. A user can record audio or manually enter a rough transcript, add context about what they are working on, and then generate a cleaned-up version that preserves meaning while fixing common transcription mistakes. The app includes a browser-based React interface, a FastAPI backend, SQLite-backed correction history, editable personal dictionary support, audio transcription endpoints, and AI-assisted correction through local Ollama or deterministic fallback rules. Voxly is also Deepgram-ready for real speech-to-text transcription and can pass vocabulary hints to improve recognition of important terms. For the hackathon demo, Voxly focuses on a practical workflow: capture rough speech, review the raw transcript, apply context-aware correction, save corrected output, and improve future results by adding important words to the personal dictionary. This makes it useful for developers, students, writers, creators, and professionals who frequently dictate notes involving specialized language. Voxly is containerized with Docker and can be run with Docker Compose, making it straightforward to test, deploy, and reproduce. The current version is intentionally focused on a reliable core experience, with future potential for richer personalization, cloud sync, team dictionaries, browser extensions, and deeper integrations with writing tools.