General stability improvements

AutoSubs-App/src-tauri/resources/AutoSubs V2.lua

@@ -10,7 +10,7 @@ print("Operating System: " .. os_name)
 -- Function to read a JSON file
 local function read_json_file(file_path)
     local file = assert(io.open(file_path, "r")) -- Open file for reading
-    local content = file:read("*a") -- Read the entire file content
+    local content = file:read("*a")              -- Read the entire file content
     file:close()
 
     -- Parse the JSON content
@@ -50,10 +50,9 @@ if os_name == "Windows" then
     -- Windows commands to open and close app using terminal commands
     command_open = 'start "" "' .. mainApp .. '"'
     command_close = 'powershell -Command "Get-Process AutoSubs | Stop-Process -Force"'
-
 elseif os_name == "OSX" then
     storagePath = os.getenv("HOME") ..
-                      "/Library/Application Support/Blackmagic Design/DaVinci Resolve/Fusion/Scripts/Utility/AutoSubs/"
+        "/Library/Application Support/Blackmagic Design/DaVinci Resolve/Fusion/Scripts/Utility/AutoSubs/"
 
     local file = assert(io.open(storagePath .. "install_path.txt", "r"))
     local install_path = file:read("*l")
@@ -79,7 +78,7 @@ local mediaPool = project:GetMediaPool()
 
 function CreateResponse(body)
     local header = "HTTP/1.1 200 OK\r\n" .. "Server: ljsocket/0.1\r\n" .. "Content-Type: application/json\r\n" ..
-                       "Content-Length: " .. #body .. "\r\n" .. "Connection: close\r\n" .. "\r\n"
+        "Content-Length: " .. #body .. "\r\n" .. "Connection: close\r\n" .. "\r\n"
 
     local response = header .. body
     return response
@@ -262,9 +261,23 @@ function ExportAudio(outputDir)
     local success, err = pcall(function()
         resolve:ImportRenderPreset(storagePath .. "render-audio-only.xml")
         project:LoadRenderPreset('render-audio-only')
+        project:SetRenderSettings({ TargetDir = outputDir })
+    end)
+
+    if not success then
+        project:LoadRenderPreset('Audio Only')
         project:SetRenderSettings({
-            TargetDir = outputDir
+            TargetDir = outputDir,
+            CustomName = "autosubs-exported-audio",
+            RenderMode = "Single clip",
+            IsExportVideo = false,
+            IsExportAudio = true,
+            AudioBitDepth = 24,
+            AudioSampleRate = 44100
         })
+    end
+
+    pcall(function()
         local pid = project:AddRenderJob()
         project:StartRendering(pid)
 
@@ -385,7 +398,7 @@ function AddSubtitles(filePath, trackIndex, templateName, textFormat, removePunc
                 end
             end
 
-            local timelineItem = mediaPool:AppendToTimeline({newClip})[1]
+            local timelineItem = mediaPool:AppendToTimeline({ newClip })[1]
 
             local subtitle = subtitles[i]
             local subtitleText = subtitle["text"]
@@ -439,7 +452,7 @@ function AddSubtitles(filePath, trackIndex, templateName, textFormat, removePunc
         end)
 
         if not success then
-            print("Error adding subtitle:", err)
+            print("Attempted to add subtitle on top of existing timeline item. Please select an empty track.")
         end
     end
 end
@@ -465,7 +478,24 @@ assert(server:set_option("nodelay", true, "tcp"))
 assert(server:set_option("reuseaddr", true))
 
 -- Bind and listen
-assert(server:bind(info))
+local success, err = pcall(function()
+    assert(server:bind(info))
+end)
+
+if not success then
+    os.execute([[
+        curl --request POST \
+            --url http://localhost:55010/ \
+            --header 'Content-Type: application/json' \
+            --header 'content-type: application/json' \
+            --data '{
+            "func":"Exit"
+        }'
+    ]])
+    sleep(0.5)
+    assert(server:bind(info))
+end
+
 assert(server:listen())
 
 -- Start AutoSubs app

AutoSubs-App/src/GlobalContext.tsx

@@ -1,7 +1,7 @@
 import { useEffect, createContext, useState, useContext, useRef } from 'react';
 import { fetch } from '@tauri-apps/plugin-http';
-import { BaseDirectory, readTextFile, exists, writeTextFile, mkdir } from '@tauri-apps/plugin-fs';
-import { join, documentDir } from '@tauri-apps/api/path';
+import { BaseDirectory, readTextFile, exists, writeTextFile } from '@tauri-apps/plugin-fs';
+import { join, documentDir, downloadDir } from '@tauri-apps/api/path';
 import { save } from '@tauri-apps/plugin-dialog';
 import { Subtitle, Speaker, TopSpeaker } from "@/types/interfaces";
 import { load, Store } from '@tauri-apps/plugin-store';
@@ -105,7 +105,7 @@ export function GlobalProvider({ children }: React.PropsWithChildren<{}>) {
 
     const [markIn, setMarkIn] = useState(0);
     const [model, setModel] = useState("small");
-    const [currentLanguage, setLanguage] = useState("en");
+    const [currentLanguage, setLanguage] = useState("");
     const [currentTemplate, setTemplate] = useState("");
     const [currentTrack, setTrack] = useState("");
     const [translate, setTranslate] = useState(false);
@@ -119,11 +119,6 @@ export function GlobalProvider({ children }: React.PropsWithChildren<{}>) {
 
     async function setTranscriptsFolder() {
         storageDir = await join(await documentDir(), "AutoSubs");
-        // create directory
-        const dirExists = await exists(storageDir, { baseDir: BaseDirectory.Document });
-        if (!dirExists) {
-            await mkdir(storageDir, { baseDir: BaseDirectory.Document, recursive: true });
-        }
     }
 
     async function getFullTranscriptPath() {
@@ -138,7 +133,7 @@ export function GlobalProvider({ children }: React.PropsWithChildren<{}>) {
 
         try {
             setModel(await store.get<string>('model') || "small");
-            setLanguage(await store.get<string>('currentLanguage') || "en");
+            setLanguage(await store.get<string>('currentLanguage') || "");
             setTemplate(await store.get<string>('currentTemplate') || "");
             setTrack(await store.get<string>('currentTrack') || "");
             setTranslate(await store.get<boolean>('translate') || false);
@@ -388,14 +383,14 @@ export function GlobalProvider({ children }: React.PropsWithChildren<{}>) {
             },
             body: JSON.stringify({
                 func: "ExportAudio",
-                outputDir: storageDir
+                outputDir: await downloadDir(),
             }),
         });
 
         const data = await response.json();
 
         if (data.timeline == "") {
-            throw new Error("You need to open a timeline in Resolve to start transcribing.");
+            throw new Error("Failed to export audio. You must have a timeline open in Resolve to start transcribing.");
         }
 
         setTimeline(data.timeline);

AutoSubs-App/src/pages/diarize-page.tsx

@@ -331,7 +331,7 @@ export function DiarizePage() {
                                 <ToggleGroup
                                     type="single"
                                     value={textFormat}
-                                    onValueChange={(value) => value && setTextFormat(value)}
+                                    onValueChange={(value: string) => value && setTextFormat(value)}
                                     className="grid grid-cols-3 gap-3 h-20"
                                 >
                                     <ToggleGroupItem

AutoSubs-App/src/pages/home-page.tsx

@@ -634,7 +634,7 @@ export function HomePage() {
                                 <ToggleGroup
                                     type="single"
                                     value={textFormat}
-                                    onValueChange={(value) => value && setTextFormat(value)}
+                                    onValueChange={(value: string) => value && setTextFormat(value)}
                                     className="grid grid-cols-3 gap-3 h-20"
                                 >
                                     <ToggleGroupItem
@@ -670,7 +670,7 @@ export function HomePage() {
                                 <Switch checked={removePunctuation} onCheckedChange={(checked) => setRemovePunctuation(checked)} />
                             </div>
                             <div className="grid gap-3">
-                                <Label htmlFor="sensitiveWords">Sensored Words</Label>
+                                <Label htmlFor="sensitiveWords">Censored Words</Label>
                                 <Input value={sensitiveWords} id="sensitiveWords" type="string" placeholder="bomb, gun, kill" onChange={(e) => setSensitiveWords(e.target.value)} />
                             </div>
                             <div className="grid grid-cols-2 gap-4">
@@ -690,7 +690,7 @@ export function HomePage() {
                                         Force Align Words
                                     </p>
                                     <p className="text-xs text-muted-foreground">
-                                        Warning: May be unstable
+                                        Improve word level timing
                                     </p>
                                 </div>
                                 <Switch checked={alignWords} onCheckedChange={(checked) => setAlignWords(checked)} />

Transcription-Server/server.py

@@ -215,17 +215,19 @@ def transcribe_audio(audio_file, kwargs, max_words, max_chars, sensitive_words):
             kwargs["model"], device=kwargs["device"], compute_type=compute_type)
         if kwargs["language"] == "auto":
             result = model.transcribe_stable(
-                audio_file, task=kwargs["task"], verbose=True, vad_filter=True, progress_callback=log_progress)
+                audio_file, task=kwargs["task"], regroup=True, verbose=True, vad_filter=True, progress_callback=log_progress)
         else:
             result = model.transcribe_stable(
-                audio_file, language=kwargs["language"], task=kwargs["task"], verbose=True, vad_filter=True, progress_callback=log_progress)
+                audio_file, language=kwargs["language"], task=kwargs["task"], regroup=True, verbose=True, vad_filter=True, progress_callback=log_progress)
             model.align(audio_file, result, kwargs["language"])
             if kwargs["align_words"]:
                 model.align_words(audio_file, result, kwargs["language"])
+                result.pad()
 
     else:
         result = stable_whisper.transcribe_any(
-            inference, audio_file, inference_kwargs=kwargs, vad=False)
+            inference, audio_file, inference_kwargs=kwargs, vad=False, regroup=True)
+        result.pad()
 
     result = modify_result(result, max_words, max_chars, sensitive_words)
 
@@ -235,10 +237,17 @@ def transcribe_audio(audio_file, kwargs, max_words, max_chars, sensitive_words):
 def diarize_audio(audio_file, device):
     from pyannote.audio import Pipeline
     print("Starting diarization...")
-    pipeline = Pipeline.from_pretrained("pyannote/speaker-diarization-3.1")
-    pipeline.to(device)
-    return pipeline(audio_file)
-
+    try:
+        pipeline = Pipeline.from_pretrained("pyannote/speaker-diarization-3.1")
+        pipeline.to(device)
+        return pipeline(audio_file)
+    except Exception as e:
+        error_message = f"failed to load diarization model. {e}"
+        print(error_message)
+        raise HTTPException(
+            status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
+            detail=error_message
+        )
 
 def merge_diarisation(transcript, diarization):
     # Array of colors to choose from
@@ -484,8 +493,13 @@ async def transcribe(request: TranscriptionRequest):
         json_filename = f"{timeline}.json"
         json_filepath = os.path.join(request.output_dir, json_filename)
         try:
+            # Create the directory if it doesn't exist
+            os.makedirs(request.output_dir, exist_ok=True)
+
+            # Save the transcription to a JSON file
             with open(json_filepath, 'w', encoding='utf-8') as f:
                 json.dump(result, f, indent=4, ensure_ascii=False)
+
             print(f"Transcription saved to: {json_filepath}")
         except Exception as e:
             print(f"Error saving JSON file: {e}")
@@ -550,10 +564,10 @@ def censor_word(result, seg_index, word_index):
 
     (
         result
-        .split_by_punctuation([('.', ' '), '。', '?', '？', ',', '，'])
-        .split_by_gap(0.4)
-        .merge_by_gap(0.1, max_words=3)
         .split_by_length(max_words=max_words, max_chars=max_chars)
+        # .split_by_punctuation([('.', ' '), '。', '?', '？', ',', '，'])
+        # .split_by_gap(0.4)
+        # .merge_by_gap(0.1, max_words=3)
     )
 
     return result