LM Studio transcription workflow still feels clunky

This is less a model problem than a workflow problem: once you have to bounce between recorder, STT engine, cleanup, and clipboard, the UX feels broken.

• –LM Studio’s docs focus on local LLMs, chat, MCP, and OpenAI-like APIs, not audio transcription, which suggests it is not really a speech front end (https://lmstudio.ai/, https://lmstudio.ai/docs/app).
• –Whisper-first tools already target this exact gap: Whisper Transcriber auto-pastes after hotkey recording, AudioWhisper does hotkey dictation plus clipboard output, and Whishper/Transcribe add local processing, diarization, and timestamps (https://www.whispertranscriber.com/, https://github.com/mazdak/AudioWhisper, https://github.com/pluja/whishper, https://icosium.org/).
• –The more interesting product is a dictation app with optional local LLM cleanup, not a generic model runner; LM Studio fits better as the backend for summarization or post-processing than as the recorder.
• –If someone nails this category, the differentiators will be hotkeys, instant paste, offline mode, and speaker-aware output more than raw transcription accuracy.