dot-emacs/local-lisp/nemo-whisper.el

;;; -*- lexical-binding: t -*-

(require 'whisper)

;; Optional: install/load posframe if you want floating display.
;; `nemo-whisper-use-posframe' controls whether it is used.
(require 'posframe nil t)

(defgroup nemo-whisper nil
  "Whisper recording utilities."
  :group 'whisper)

(defcustom nemo-whisper-use-posframe t
  "Display the Whisper buffer in a floating posframe when non-nil."
  :type 'boolean
  :group 'nemo-whisper)

(defcustom nemo-whisper-hide-posframe-after-transcription t
  "Hide the posframe after copying the transcription to the kill ring."
  :type 'boolean
  :group 'nemo-whisper)

(defcustom nemo-whisper-posframe-width 80
  "Width of the Whisper posframe."
  :type 'integer
  :group 'nemo-whisper)

(defcustom nemo-whisper-posframe-height 8
  "Height of the Whisper posframe."
  :type 'integer
  :group 'nemo-whisper)

(defvar nemo-whisper--buffer nil
  "Buffer currently used for Whisper recording.")

(defvar nemo-whisper--previous-buffer nil
  "Buffer that was selected before Whisper capture started.")

(defvar nemo-whisper--recording-p nil
  "Whether Whisper is currently recording.")

(defvar nemo-whisper--transcribing-p nil
  "Whether Whisper is currently transcribing.")

(defvar nemo-whisper--posframe-active-p nil
  "Whether the current capture buffer is displayed in a posframe.")


(defvar nemo-whisper--capture-map
  (let ((map (make-sparse-keymap)))
    ;; Bind both forms because terminal and GUI Emacs can report Escape
    ;; differently.
    (define-key map (kbd "ESC") #'nemo-whisper-escape)
    (define-key map [escape] #'nemo-whisper-escape)
    map)
  "Keymap for a Whisper capture buffer.")

(define-minor-mode nemo-whisper-capture-mode
  "Minor mode used by the Whisper capture buffer."
  :init-value nil
  :lighter " Whisper"
  :keymap nemo-whisper--capture-map)

(defun nemo-whisper--display-buffer (buffer)
  "Display BUFFER normally or in a posframe."
  (if (and nemo-whisper-use-posframe
           (fboundp 'posframe-show)
           (display-graphic-p))
      (progn
        (setq nemo-whisper--posframe-active-p t)

        (posframe-show
         buffer
         :position nil
         :poshandler #'posframe-poshandler-frame-center
         :width nemo-whisper-posframe-width
         :height nemo-whisper-posframe-height
         :border-width 2
         :internal-border-width 8
         :internal-border-color "gray50"
         :accept-focus t)

        ;; Explicitly focus the child frame.  Without this, keyboard input
        ;; may continue going to the previously selected Emacs frame.
        (let ((window (get-buffer-window buffer t)))
          (when (window-live-p window)
            (select-frame-set-input-focus (window-frame window))
            (select-window window))))

    (setq nemo-whisper--posframe-active-p nil)
    (pop-to-buffer buffer)))

(defun nemo-whisper--remove-recording-preview ()
  "Remove the recording preview before Whisper inserts its result."
  (let ((inhibit-read-only t))
    (erase-buffer)
    (goto-char (point-min))))

(defun nemo-whisper--stop-recording ()
  "Stop recording and start Whisper transcription."
  (when (and nemo-whisper--buffer
             (buffer-live-p nemo-whisper--buffer)
             nemo-whisper--recording-p)
    (setq nemo-whisper--recording-p nil
          nemo-whisper--transcribing-p t)

    (with-current-buffer nemo-whisper--buffer
      ;; Remove “Recording...” so the transcription is the only text
      ;; inserted into the buffer.
      (nemo-whisper--remove-recording-preview)
      (whisper-run))

    (message "Recording stopped; transcribing...")))

(defun nemo-whisper--copy-result-to-kill-ring ()
  "Copy the completed transcription to the kill ring."
  (let ((text (string-trim
               (buffer-substring-no-properties
                (point-min)
                (point-max)))))
    (unless (string-empty-p text)
      (kill-new text)
      (message "Whisper transcription copied to the kill ring"))

    (setq nemo-whisper--transcribing-p nil)

    ;; Select the completed transcription.  This makes a subsequent ESC
    ;; return to the previous buffer when the capture buffer remains visible.
    (goto-char (point-min))
    (push-mark (point-max) t t)

    (when (and nemo-whisper--posframe-active-p
               nemo-whisper-hide-posframe-after-transcription
               (fboundp 'posframe-hide))
      (posframe-hide (buffer-name)))))

(defun nemo-whisper--return-to-previous-buffer ()
  "Return to the buffer active before Whisper capture started."
  (when (and nemo-whisper--posframe-active-p
             (fboundp 'posframe-hide)
             nemo-whisper--buffer)
    (posframe-hide (buffer-name nemo-whisper--buffer)))

  (let ((previous-buffer nemo-whisper--previous-buffer))
    (setq nemo-whisper--buffer nil
          nemo-whisper--previous-buffer nil
          nemo-whisper--recording-p nil
          nemo-whisper--transcribing-p nil
          nemo-whisper--posframe-active-p nil)

    (when (buffer-live-p previous-buffer)
      (pop-to-buffer previous-buffer))))

(defun nemo-whisper-escape ()
  "Stop recording, or return to the previous buffer after transcription."
  (interactive)
  (cond
   (nemo-whisper--recording-p
    (nemo-whisper--stop-recording))

   (nemo-whisper--transcribing-p
    (message "Whisper is still transcribing; please wait"))

   (t
    (nemo-whisper--return-to-previous-buffer))))

(defun nemo-whisper-record-to-kill-ring ()
  "Record speech and copy the transcription to the kill ring.

Invoke this command to create a capture buffer and start recording.
Press ESC to stop recording.  After transcription completes, press ESC
again to return to the previously selected buffer."
  (interactive)
  (cond
   ;; If called while recording, treat the command as a stop command.
   (nemo-whisper--recording-p
    (nemo-whisper--stop-recording))

   ;; Do not interrupt an active transcription.
   (nemo-whisper--transcribing-p
    (message "Whisper is still transcribing; please wait"))

   ;; Start a new capture.
   (t
    (setq nemo-whisper--previous-buffer
          (window-buffer (selected-window))
          nemo-whisper--buffer
          (generate-new-buffer "*Whisper capture*")
          nemo-whisper--recording-p t
          nemo-whisper--transcribing-p nil)

    (with-current-buffer nemo-whisper--buffer
      (text-mode)
      (nemo-whisper-capture-mode 1)

      ;; Insert the temporary preview text.
      (insert (propertize "Recording...\n"
                          'face 'font-lock-warning-face))

      ;; Ask Whisper to insert its result into this buffer.
      (setq-local whisper-insert-text-at-point t)

      ;; Prevent Whisper from opening another transcription buffer.
      (setq-local whisper-display-transcription-buffer nil)

      ;; This hook is buffer-local, so it will not run in Whisper's
      ;; separate transcription buffer.
      (add-hook 'whisper-after-insert-hook
                #'nemo-whisper--copy-result-to-kill-ring
                nil
                t))

    (nemo-whisper--display-buffer nemo-whisper--buffer)

    ;; The first invocation starts recording.
    (with-current-buffer nemo-whisper--buffer
      (whisper-run))

    (message "Recording started; press ESC when finished speaking"))))

(provide 'nemo-whisper)