feat: voice to text

config/hypr/hyprland.conf

@@ -135,6 +135,7 @@ bind = , I, exec, hyprctl dispatch submap reset; ctl idle
 bind = , K, exec, hyprctl dispatch submap reset; ctl keyboard next
 bind = , M, exec, hyprctl dispatch submap reset; ctl media
 bind = , P, exec, hyprctl dispatch submap reset; ctl power
+bind = , S, exec, hyprctl dispatch submap reset; ctl dictate
 bind = , T, exec, hyprctl dispatch submap reset; theme
 
 bind = , catchall, submap, reset

flake.lock

@@ -39,24 +39,6 @@
         "type": "github"
       }
     },
-    "flake-utils": {
-      "inputs": {
-        "systems": "systems"
-      },
-      "locked": {
-        "lastModified": 1731533236,
-        "narHash": "sha256-l0KFg5HjrsfsO/JpG+r7fRrqm12kzFHyUHqHCVpMMbI=",
-        "owner": "numtide",
-        "repo": "flake-utils",
-        "rev": "11707dc2f618dd54ca8739b309ec4fc024de578b",
-        "type": "github"
-      },
-      "original": {
-        "owner": "numtide",
-        "repo": "flake-utils",
-        "type": "github"
-      }
-    },
     "home-manager": {
       "inputs": {
         "nixpkgs": [
@@ -220,46 +202,9 @@
         "neovim-nightly": "neovim-nightly",
         "nixos-hardware": "nixos-hardware",
         "nixpkgs": "nixpkgs_3",
-        "whisper-dictation": "whisper-dictation",
         "zen-browser": "zen-browser"
       }
     },
-    "systems": {
-      "locked": {
-        "lastModified": 1681028828,
-        "narHash": "sha256-Vy1rq5AaRuLzOxct8nz4T6wlgyUR7zLU309k9mBC768=",
-        "owner": "nix-systems",
-        "repo": "default",
-        "rev": "da67096a3b9bf56a91d16901293e51ba5b49a27e",
-        "type": "github"
-      },
-      "original": {
-        "owner": "nix-systems",
-        "repo": "default",
-        "type": "github"
-      }
-    },
-    "whisper-dictation": {
-      "inputs": {
-        "flake-utils": "flake-utils",
-        "nixpkgs": [
-          "nixpkgs"
-        ]
-      },
-      "locked": {
-        "lastModified": 1771198514,
-        "narHash": "sha256-5Dd0zVTh+nUf3lxrDpgQSHbr0pU40hGh/hUuRbwG790=",
-        "owner": "jacopone",
-        "repo": "whisper-dictation",
-        "rev": "20190a148d8bce6f0d0b78a56cc22afe02844e8f",
-        "type": "github"
-      },
-      "original": {
-        "owner": "jacopone",
-        "repo": "whisper-dictation",
-        "type": "github"
-      }
-    },
     "zen-browser": {
       "inputs": {
         "home-manager": "home-manager_2",

flake.nix

@@ -11,10 +11,6 @@
     zen-browser.url = "github:0xc000022070/zen-browser-flake";
     claude-code.url = "github:ryoppippi/claude-code-overlay";
     neovim-nightly.url = "github:nix-community/neovim-nightly-overlay";
-    whisper-dictation = {
-      url = "github:jacopone/whisper-dictation";
-      inputs.nixpkgs.follows = "nixpkgs";
-    };
 };
 
   outputs =
@@ -25,7 +21,6 @@
       zen-browser,
       claude-code,
       neovim-nightly,
-      whisper-dictation,
       ...
     }:
     let
@@ -92,7 +87,7 @@
         home-manager.lib.homeManagerConfiguration {
           pkgs = mkPkgs hostConfig.platform [ ];
           extraSpecialArgs = {
-            inherit zen-browser whisper-dictation hostConfig;
+            inherit zen-browser hostConfig;
           };
           modules = [ ./home/home.nix ];
         };
@@ -134,6 +129,10 @@
                   "tailscale"
                   "libfprint-2-tod1-goodix"
                   "brgenml1lpr"
+                  "cuda_cccl"
+                  "cuda_cudart"
+                  "libcublas"
+                  "cuda_nvcc"
                 ]
               );
           }
@@ -144,7 +143,7 @@
             home-manager.backupFileExtension = "bak";
             home-manager.users.barrett = import ./home/home.nix;
             home-manager.extraSpecialArgs = {
-              inherit zen-browser whisper-dictation;
+              inherit zen-browser;
               hostConfig = xps15Config;
             };
           }

home/modules/dictation.nix

@@ -3,19 +3,17 @@
   pkgs,
   config,
   hostConfig,
-  whisper-dictation,
   ...
 }:
 
 let
+  whisper = pkgs.whisper-cpp.override { cudaSupport = hostConfig.gpu == "nvidia"; };
   modelDir = "${config.home.homeDirectory}/.local/share/whisper-models";
-  model = "ggml-base.bin";
+  model = "ggml-medium.bin";
   modelUrl = "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/${model}";
 in
 {
-  home.packages = [
+  home.packages = [ whisper ];
-    whisper-dictation.packages.${hostConfig.platform}.default
-  ];
 
   home.activation.downloadWhisperModel = lib.hm.dag.entryAfter [ "writeBoundary" ] ''
     if [ ! -f "${modelDir}/${model}" ]; then
@@ -23,29 +21,4 @@ in
       run ${pkgs.curl}/bin/curl -L -o "${modelDir}/${model}" "${modelUrl}"
     fi
   '';
-
-  xdg.configFile."whisper-dictation/config.yaml".text = builtins.toJSON {
-    whisper = {
-      model = "base";
-      language = "auto";
-    };
-    hotkey = {
-      key = "KEY_DOT";
-      modifiers = [ "KEY_LEFTMETA" ];
-    };
-  };
-
-  systemd.user.services.whisper-dictation = {
-    Unit = {
-      Description = "Whisper Dictation speech-to-text daemon";
-      After = [ "graphical-session.target" "ydotoold.service" ];
-      PartOf = [ "graphical-session.target" ];
-    };
-    Service = {
-      ExecStart = "${whisper-dictation.packages.${hostConfig.platform}.default}/bin/whisper-dictation";
-      Restart = "on-failure";
-      RestartSec = 5;
-    };
-    Install.WantedBy = [ "graphical-session.target" ];
-  };
 }

hosts/xps15/configuration.nix

@@ -95,24 +95,10 @@ in
       "libvirt"
       "storage"
       "power"
-      "input"
     ];
     shell = pkgs.zsh;
   };
 
-  hardware.uinput.enable = true;
-
-  systemd.user.services.ydotoold = {
-    description = "ydotool daemon";
-    wantedBy = [ "graphical-session.target" ];
-    partOf = [ "graphical-session.target" ];
-    serviceConfig = {
-      ExecStart = "${pkgs.ydotool}/bin/ydotoold";
-      Restart = "on-failure";
-      RestartSec = 3;
-    };
-  };
-
   programs.chromium = {
     enable = true;
     extraOpts = {

scripts/ctl

@@ -533,6 +533,51 @@ power)
   "$shutdown") systemctl poweroff ;;
   esac
   ;;
+dictate)
+  require pw-record whisper-cli wl-copy notify-send
+  dtmp="${XDG_RUNTIME_DIR:-/tmp}/dictation"
+  mkdir -p "$dtmp"
+  dpid="$dtmp/rec_pid"
+  daudio="$dtmp/recording.wav"
+  dmodel="${DICTATE_MODEL:-medium}"
+  dmodel_dir="${XDG_DATA_HOME:-$HOME/.local/share}/whisper-models"
+  dmodel_file="$dmodel_dir/ggml-$dmodel.bin"
+
+  if [ -f "$dpid" ] && kill -0 "$(cat "$dpid")" 2>/dev/null; then
+    kill "$(cat "$dpid")" 2>/dev/null
+    rm -f "$dpid"
+    sleep 0.2
+    if [ ! -s "$daudio" ]; then
+      notify-send -a ctl "no audio"
+      exit 1
+    fi
+    notify-send -a ctl "transcribing..."
+    text=$(whisper-cli \
+      --model "$dmodel_file" \
+      --language "${DICTATE_LANG:-en}" \
+      --no-prints --no-timestamps \
+      "$daudio" 2>/dev/null)
+    rm -f "$daudio"
+    text=$(printf '%s' "$text" | sed 's/^[[:space:]]*//;s/[[:space:]]*$//' | tr -s ' ')
+    if [ -z "$text" ]; then
+      notify-send -a ctl "no speech detected"
+      exit 1
+    fi
+    printf '%s' "$text" | wl-copy
+    notify-send -a ctl "$text"
+    exit 0
+  fi
+
+  if [ ! -f "$dmodel_file" ]; then
+    notify-send -a ctl "downloading whisper $dmodel model..."
+    mkdir -p "$dmodel_dir"
+    curl -L -o "$dmodel_file" \
+      "https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-$dmodel.bin"
+  fi
+  notify-send -a ctl "recording..."
+  pw-record "$daudio" &
+  printf '%s' "$!" > "$dpid"
+  ;;
 idle)
   require notify-send
   if systemctl --user is-active --quiet hypridle.service; then
@@ -546,7 +591,7 @@ idle)
   fi
   ;;
 *)
-  echo "Usage: ctl {screenshot|keyboard|audio|wifi|brightness|volume|media|wallpaper|power|idle|clip}" >&2
+  echo "Usage: ctl {screenshot|keyboard|audio|wifi|brightness|volume|media|wallpaper|power|idle|clip|dictate}" >&2
   exit 1
   ;;
 esac