feat(transcription): add word correction with phonetic matching

src-tauri/Cargo.lock

@@ -1105,7 +1105,7 @@ dependencies = [
  "libc",
  "option-ext",
  "redox_users 0.5.0",
- "windows-sys 0.59.0",
+ "windows-sys 0.60.2",
 ]
 
 [[package]]
@@ -2001,6 +2001,7 @@ dependencies = [
  "futures-util",
  "hound",
  "log",
+ "natural",
  "once_cell",
  "ort-sys",
  "rdev",
@@ -2010,6 +2011,7 @@ dependencies = [
  "rustfft",
  "serde",
  "serde_json",
+ "strsim",
  "tauri",
  "tauri-build",
  "tauri-plugin-autostart",
@@ -2962,6 +2964,15 @@ dependencies = [
  "tempfile",
 ]
 
+[[package]]
+name = "natural"
+version = "0.5.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "65e6b44f8ddc659cde3555e0140d3441ad26cb03a1410774af1f9a19097c1867"
+dependencies = [
+ "rust-stemmers",
+]
+
 [[package]]
 name = "ndarray"
 version = "0.16.1"
@@ -4365,6 +4376,16 @@ dependencies = [
  "realfft",
 ]
 
+[[package]]
+name = "rust-stemmers"
+version = "1.2.0"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "e46a2036019fdb888131db7a4c847a1063a7493f971ed94ea82c67eada63ca54"
+dependencies = [
+ "serde",
+ "serde_derive",
+]
+
 [[package]]
 name = "rustc-demangle"
 version = "0.1.24"

src-tauri/Cargo.toml

@@ -57,6 +57,8 @@ reqwest = { version = "0.11", features = ["json", "stream"] }
 futures-util = "0.3"
 tauri-plugin-fs = "2"
 rustfft = "6.4.0"
+strsim = "0.11.0"
+natural = "0.5.0"
 
 [dependencies.ort-sys]
 version = "=2.0.0-rc.9"

src-tauri/src/lib.rs

@@ -187,6 +187,7 @@ pub fn run() {
             shortcut::change_selected_language_setting,
             shortcut::change_overlay_position_setting,
             shortcut::change_debug_mode_setting,
+            shortcut::update_correct_words,
             shortcut::suspend_binding,
             shortcut::resume_binding,
             trigger_update_check,

src-tauri/src/managers/transcription.rs

@@ -1,8 +1,10 @@
 use crate::managers::model::ModelManager;
 use crate::settings::get_settings;
 use anyhow::Result;
+use natural::phonetics::soundex;
 use serde::Serialize;
 use std::sync::{Arc, Mutex};
+use strsim::levenshtein;
 use tauri::{App, AppHandle, Emitter, Manager};
 use whisper_rs::install_whisper_log_trampoline;
 use whisper_rs::{
@@ -25,6 +27,82 @@ pub struct TranscriptionManager {
     current_model_id: Mutex<Option<String>>,
 }
 
+fn correct_words(text: &str, correct_words: &[String]) -> String {
+    if correct_words.is_empty() {
+        return text.to_string();
+    }
+
+    let words: Vec<&str> = text.split_whitespace().collect();
+    let mut corrected_words = Vec::new();
+
+    for word in words {
+        let cleaned_word = word.trim_matches(|c: char| !c.is_alphabetic()).to_lowercase();
+        
+        if cleaned_word.is_empty() {
+            corrected_words.push(word.to_string());
+            continue;
+        }
+
+        let mut best_match: Option<&String> = None;
+        let mut best_score = f64::MAX;
+
+        for correct_word in correct_words {
+            let correct_word_lower = correct_word.to_lowercase();
+            
+            // Calculate Levenshtein distance (normalized by length)
+            let levenshtein_dist = levenshtein(&cleaned_word, &correct_word_lower);
+            let max_len = cleaned_word.len().max(correct_word_lower.len()) as f64;
+            let levenshtein_score = if max_len > 0.0 { levenshtein_dist as f64 / max_len } else { 1.0 };
+            
+            // Calculate phonetic similarity using Soundex
+            let phonetic_match = soundex(&cleaned_word, &correct_word_lower);
+            
+            // Combine scores: favor phonetic matches, but also consider string similarity
+            let combined_score = if phonetic_match {
+                levenshtein_score * 0.3  // Give significant boost to phonetic matches
+            } else {
+                levenshtein_score
+            };
+            
+            // Accept if the score is good enough (threshold: 0.4 for good matches)
+            if combined_score < 0.4 && combined_score < best_score {
+                best_match = Some(correct_word);
+                best_score = combined_score;
+            }
+        }
+
+        if let Some(replacement) = best_match {
+            // Preserve the original case pattern as much as possible
+            let corrected = if word.chars().all(|c| c.is_uppercase()) {
+                replacement.to_uppercase()
+            } else if word.chars().next().map_or(false, |c| c.is_uppercase()) {
+                let mut chars: Vec<char> = replacement.chars().collect();
+                if let Some(first_char) = chars.get_mut(0) {
+                    *first_char = first_char.to_uppercase().next().unwrap_or(*first_char);
+                }
+                chars.into_iter().collect()
+            } else {
+                replacement.clone()
+            };
+            
+            // Preserve punctuation from original word
+            let original_suffix: String = word.chars().rev()
+                .take_while(|c| !c.is_alphabetic())
+                .collect::<String>()
+                .chars().rev().collect();
+            let original_prefix: String = word.chars()
+                .take_while(|c| !c.is_alphabetic())
+                .collect();
+                
+            corrected_words.push(format!("{}{}{}", original_prefix, corrected, original_suffix));
+        } else {
+            corrected_words.push(word.to_string());
+        }
+    }
+
+    corrected_words.join(" ")
+}
+
 impl TranscriptionManager {
     pub fn new(app: &App, model_manager: Arc<ModelManager>) -> Result<Self> {
         let app_handle = app.app_handle().clone();
@@ -207,6 +285,13 @@ impl TranscriptionManager {
             result.push_str(&segment);
         }
 
+        // Apply word correction if correct words are configured
+        let corrected_result = if !settings.correct_words.is_empty() {
+            correct_words(&result, &settings.correct_words)
+        } else {
+            result
+        };
+
         let et = std::time::Instant::now();
         let translation_note = if settings.translate_to_english {
             " (translated)"
@@ -215,6 +300,6 @@ impl TranscriptionManager {
         };
         println!("\ntook {}ms{}", (et - st).as_millis(), translation_note);
 
-        Ok(result.trim().to_string())
+        Ok(corrected_result.trim().to_string())
     }
 }

src-tauri/src/settings.rs

@@ -42,6 +42,8 @@ pub struct AppSettings {
     pub overlay_position: OverlayPosition,
     #[serde(default = "default_debug_mode")]
     pub debug_mode: bool,
+    #[serde(default)]
+    pub correct_words: Vec<String>,
 }
 
 fn default_model() -> String {
@@ -104,6 +106,7 @@ pub fn get_default_settings() -> AppSettings {
         selected_language: "auto".to_string(),
         overlay_position: OverlayPosition::Bottom,
         debug_mode: false,
+        correct_words: Vec::new(),
     }
 }
 

src-tauri/src/shortcut.rs

@@ -163,6 +163,14 @@ pub fn change_debug_mode_setting(app: AppHandle, enabled: bool) -> Result<(), St
     Ok(())
 }
 
+#[tauri::command]
+pub fn update_correct_words(app: AppHandle, words: Vec<String>) -> Result<(), String> {
+    let mut settings = settings::get_settings(&app);
+    settings.correct_words = words;
+    settings::write_settings(&app, settings);
+    Ok(())
+}
+
 /// Determine whether a shortcut string contains at least one non-modifier key.
 /// We allow single non-modifier keys (e.g. "f5" or "space") but disallow
 /// modifier-only combos (e.g. "ctrl" or "ctrl+shift").

src-tauri/tauri.conf.json

@@ -15,9 +15,9 @@
       {
         "title": "Handy",
         "width": 540,
-        "height": 700,
+        "height": 750,
         "minWidth": 540,
-        "minHeight": 700,
+        "minHeight": 750,
         "resizable": true,
         "maximizable": false
       }

src/App.tsx

@@ -1,12 +1,12 @@
 import { invoke } from "@tauri-apps/api/core";
 import { useEffect, useState } from "react";
+import { Toaster } from "sonner";
 import "./App.css";
 import AccessibilityPermissions from "./components/AccessibilityPermissions";
 import Footer from "./components/footer";
 import HandyTextLogo from "./components/icons/HandyTextLogo";
 import Onboarding from "./components/onboarding";
 import { Settings } from "./components/settings/Settings";
-import { Toaster } from "sonner";
 
 function App() {
   const [showOnboarding, setShowOnboarding] = useState<boolean | null>(null);
@@ -35,7 +35,7 @@ function App() {
     return (
       <div className="min-h-screen flex flex-col w-full">
         <div className="flex flex-col items-center pt-6 gap-8 px-4 flex-1">
-          <HandyTextLogo width={240} />
+          <HandyTextLogo width={200} />
           <Onboarding onModelSelected={handleModelSelected} />
         </div>
       </div>
@@ -46,7 +46,7 @@ function App() {
     <div className="min-h-screen flex flex-col w-full">
       <Toaster />
       <div className="flex flex-col items-center pt-6 gap-8 px-4 flex-1">
-        <HandyTextLogo width={240} />
+        <HandyTextLogo width={200} />
         <AccessibilityPermissions />
         <Settings />
       </div>

src/components/settings/CorrectWords.tsx

@@ -0,0 +1,113 @@
+import React, { useState } from "react";
+import { SettingContainer } from "../ui/SettingContainer";
+import { useSettings } from "../../hooks/useSettings";
+
+interface CorrectWordsProps {
+  descriptionMode?: "inline" | "tooltip";
+  grouped?: boolean;
+}
+
+export const CorrectWords: React.FC<CorrectWordsProps> = ({
+  descriptionMode = "tooltip",
+  grouped = false,
+}) => {
+  const { getSetting, updateSetting, isUpdating } = useSettings();
+  const [newWord, setNewWord] = useState("");
+
+  const correctWords = getSetting("correct_words") || [];
+
+  const handleAddWord = () => {
+    const trimmedWord = newWord.trim();
+    // Don't allow spaces in words
+    if (trimmedWord && !trimmedWord.includes(' ') && !correctWords.includes(trimmedWord)) {
+      updateSetting("correct_words", [...correctWords, trimmedWord]);
+      setNewWord("");
+    }
+  };
+
+  const handleRemoveWord = (wordToRemove: string) => {
+    updateSetting(
+      "correct_words", 
+      correctWords.filter((word) => word !== wordToRemove)
+    );
+  };
+
+  const handleKeyPress = (e: React.KeyboardEvent) => {
+    if (e.key === "Enter") {
+      e.preventDefault();
+      handleAddWord();
+    }
+  };
+
+  return (
+    <div className={`px-4 p-2 ${grouped ? "" : "rounded-lg border border-mid-gray/20"}`}>
+      {/* Title, info button, input and add button all in one line */}
+      <div className="flex items-center gap-2">
+        <h3 className="text-sm font-medium">Custom Words</h3>
+        {descriptionMode === "tooltip" && (
+          <div title="Add words that are often misheard or misspelled during transcription. The system will automatically correct similar-sounding words to match your list.">
+            <svg
+              className="w-4 h-4 text-mid-gray cursor-help hover:text-logo-primary transition-colors duration-200 select-none"
+              fill="none"
+              stroke="currentColor"
+              viewBox="0 0 24 24"
+            >
+              <path
+                strokeLinecap="round"
+                strokeLinejoin="round"
+                strokeWidth={2}
+                d="M13 16h-1v-4h-1m1-4h.01M21 12a9 9 0 11-18 0 9 9 0 0118 0z"
+              />
+            </svg>
+          </div>
+        )}
+        <input
+          type="text"
+          value={newWord}
+          onChange={(e) => setNewWord(e.target.value)}
+          onKeyPress={handleKeyPress}
+          placeholder="Add a word"
+          className="flex-1 px-2 py-1 text-sm border border-mid-gray/30 rounded focus:outline-none focus:ring-1 focus:ring-logo-primary focus:border-transparent bg-background"
+          disabled={isUpdating("correct_words")}
+        />
+        <button
+          onClick={handleAddWord}
+          disabled={!newWord.trim() || newWord.includes(' ') || isUpdating("correct_words")}
+          className="px-3 py-1 text-xs font-medium text-white bg-logo-primary rounded hover:bg-logo-primary/90 focus:outline-none focus:ring-1 focus:ring-logo-primary disabled:opacity-50 disabled:cursor-not-allowed transition-colors"
+        >
+          Add
+        </button>
+      </div>
+
+      {/* Words list - flex wrap layout */}
+      {correctWords.length > 0 && (
+        <div className="mt-2 flex flex-wrap gap-1">
+          {correctWords.map((word) => (
+            <button
+              key={word}
+              onClick={() => handleRemoveWord(word)}
+              disabled={isUpdating("correct_words")}
+              className="inline-flex items-center gap-1 px-2 py-1 bg-mid-gray/10 rounded text-xs hover:bg-red-100 hover:text-red-700 focus:outline-none disabled:opacity-50 cursor-pointer transition-colors"
+              aria-label={`Remove ${word}`}
+            >
+              <span>{word}</span>
+              <svg
+                className="w-3 h-3"
+                fill="none"
+                stroke="currentColor"
+                viewBox="0 0 24 24"
+              >
+                <path
+                  strokeLinecap="round"
+                  strokeLinejoin="round"
+                  strokeWidth={3}
+                  d="M6 18L18 6M6 6l12 12"
+                />
+              </svg>
+            </button>
+          ))}
+        </div>
+      )}
+    </div>
+  );
+};

src/components/settings/Settings.tsx

@@ -8,6 +8,7 @@ import { ShowOverlay } from "./ShowOverlay";
 import { HandyShortcut } from "./HandyShortcut";
 import { TranslateToEnglish } from "./TranslateToEnglish";
 import { LanguageSelector } from "./LanguageSelector";
+import { CorrectWords } from "./CorrectWords";
 import { SettingsGroup } from "../ui/SettingsGroup";
 import { DebugSettings } from "./debug";
 import { useSettings } from "../../hooks/useSettings";
@@ -54,6 +55,7 @@ export const Settings: React.FC = () => {
         <OutputDeviceSelector descriptionMode="tooltip" grouped={true} />
         <ShowOverlay descriptionMode="tooltip" grouped={true} />
         <TranslateToEnglish descriptionMode="tooltip" grouped={true} />
+        <CorrectWords descriptionMode="tooltip" grouped={true} />
         <AlwaysOnMicrophone descriptionMode="tooltip" grouped={true} />
       </SettingsGroup>
 

src/components/settings/index.ts

@@ -7,3 +7,4 @@ export { AudioFeedback } from "./AudioFeedback";
 export { ShowOverlay } from "./ShowOverlay";
 export { HandyShortcut } from "./HandyShortcut";
 export { TranslateToEnglish } from "./TranslateToEnglish";
+export { CorrectWords } from "./CorrectWords";

src/hooks/useSettings.ts

@@ -205,6 +205,9 @@ export const useSettings = (): UseSettingsReturn => {
           case "debug_mode":
             await invoke("change_debug_mode_setting", { enabled: value });
             break;
+          case "correct_words":
+            await invoke("update_correct_words", { words: value });
+            break;
           case "bindings":
             // Handle bindings separately - they use their own invoke methods
             break;
@@ -251,6 +254,7 @@ export const useSettings = (): UseSettingsReturn => {
         selected_language: "auto",
         overlay_position: "bottom",
         debug_mode: false,
+        correct_words: [],
       };
 
       const defaultValue = defaults[key];

src/lib/types.ts

@@ -34,6 +34,7 @@ export const SettingsSchema = z.object({
   selected_language: z.string(),
   overlay_position: OverlayPositionSchema,
   debug_mode: z.boolean(),
+  correct_words: z.array(z.string()).optional().default([]),
 });
 
 export const BindingResponseSchema = z.object({