blairjordan/pasteface is a Rust-based tool that records and transcribes speech in the terminal so you can work with it as text. The latest commit 5b32a22(2026-10-29) changes the control display on the speech screen to match the screen state, and limits the shimmer of the in-progress display to the generating portion only.
Previously, the play, voice-select, and stop keys always appeared together on the bottom line, and each key's click target was fixed at columns 11 and 23. Now, while playback is in progress the primary button shows a stop display, and in the idle state it shows a play display. The button width is computed from the label length so that the click target and the on-screen display no longer diverge, and the voice button's position was moved to match that width. 'Esc stops' was removed from the status text, and the initial guidance text and the playback-ended text were also shortened.
The newly added speech_status function splits the status text on ' · ' and renders it piece by piece, applying the shimmer only to pieces that begin with 'Generating' or 'generating'. The elapsed time is appended and shown only while generating. During playback a line such as 'Playing part 5/61 · generating 14/61' appears as is, so on screen you can tell which part is still being generated.
The commit also includes a unit test that draws the screen with ratatui's TestBackend to check that only the play display remains in the idle state and that the stop display appears in the playback state. That said, this test is limited to the on-screen display. The commit information contains no release version, performance figures, or user metrics, and the scope of the change is narrow, confined to the single file src/speech.rs. Since this is not a new feature but a refinement of the speech screen, for people who were looking for a recording/transcription tool itself it is close to a change worth checking again.