Different input, same promise. Wispr Flow turns speech into polished text in any app through a closed cloud pipeline. OpenCues works on typed text: you place a _ and the model fills it, rewrites in place, or pulls live data, on whichever provider you choose.
Where they overlap
Flow's Command Mode (speak an instruction over selected text and it rewrites in place) is the closest thing to OpenCues transform blanks. Everything else is complementary: Flow is dictation, OpenCues is computation at the cursor.
Where Wispr Flow struggles
Command Mode is clipboard engineering: your selection is captured (accessibility read or a synthetic copy), rewritten in Flow's cloud, and pasted back over the still-active selection. That means edits are wholesale swaps with no model of your text, so revert is whatever the host app's undo does, and nonstandard editors degrade to fallbacks. Audio always leaves the machine: transcription is cloud-only with no offline mode, and Command Mode itself is a paid, opt-in feature.
By the numbers
By the measured flows: two steps to speak an answer into place (hold, speak) against OpenCues' one typed step, three for a spoken rewrite against one. Both land with zero end-lag. The difference is recovery: keeping part of a Flow rewrite means re-editing by hand, where OpenCues reverts any single word in two keys. Equal-ish steps, different range: those two held-key steps carry dictation and spoken commands, while the one typed step also summons live data, settings, and per-word revertable transforms. Sequences in the full comparison.
How OpenCues differs
OpenCues is Apache-2.0 open source and free; Flow is proprietary at $15/mo. Flow transcribes in its own cloud with no local mode or bring-your-own keys; OpenCues runs on ten providers, your existing Claude or ChatGPT plan, or fully offline via Ollama. OpenCues does not take voice input, so the honest pairing is Flow putting words in the field and OpenCues working on whatever is there.
OpenCues also keeps PII off providers (identity values are tokenised before any cloud call) and its edits are revertable per word. Flow answers back with reach OpenCues does not have: phones and native desktop apps.
OpenCues is also extensible by design: cues and blanks are markdown files you write yourself, and blanks can bind shell scripts or runtime classes, all under an open standard anyone can implement. See write your own cue and write your own blank.
Related: Full comparison · LLM providers · Wispr Flow