updated app

app.py

@@ -62,7 +62,7 @@ load_model()
 def generate_tokens(image, user_prompt):
     # Same prompt preparation (unchanged)
     messages = [
-        {"role": "system", "content": "You are a helpful assistant. If you see a person in the image, this is me interacting, so if the question involve me as a personal pronoun, answer it. Always answer very briefly and in one sentence while being concise, using only essential information."},
+        {"role": "system", "content": "You are a helpful assistant. If you see a person in the image, this is me interacting, so if the question involve me as a personal, answer it based on my actions. Always answer very briefly and in one sentence while being concise, using only essential information."},
         {"role": "user", "content": f"<image>\n{user_prompt}"}
     ]
     rendered = tokenizer.apply_chat_template(

static/app.js

@@ -105,35 +105,32 @@ function setupVoicer() {
     });
 }
 
-// Draw video frame to canvas (always call RAF to keep loop alive; skip draw if generating)
+// Draw video frame to canvas (always draw, no pause during generation)
 function drawVideo() {
     const width = canvasElement.width;
     const height = canvasElement.height;
-
-    if (!isGenerating) {
-        canvasCtx.clearRect(0, 0, width, height);
-
-        if (video.readyState === video.HAVE_ENOUGH_DATA && video.videoWidth > 0 && video.videoHeight > 0) {
-            const videoAspect = video.videoWidth / video.videoHeight;
-            const canvasAspect = width / height;
-            let drawWidth, drawHeight, offsetX, offsetY;
-
-            if (videoAspect > canvasAspect) {
-                drawHeight = height;
-                drawWidth = height * videoAspect;
-                offsetX = (width - drawWidth) / 2;
-                offsetY = 0;
-            } else {
-                drawWidth = width;
-                drawHeight = width / videoAspect;
-                offsetX = 0;
-                offsetY = (height - drawHeight) / 2;
-            }
-
-            canvasCtx.drawImage(video, offsetX, offsetY, drawWidth, drawHeight);
+    canvasCtx.clearRect(0, 0, width, height);
+
+    if (video.readyState === video.HAVE_ENOUGH_DATA && video.videoWidth > 0 && video.videoHeight > 0) {
+        const videoAspect = video.videoWidth / video.videoHeight;
+        const canvasAspect = width / height;
+        let drawWidth, drawHeight, offsetX, offsetY;
+
+        if (videoAspect > canvasAspect) {
+            drawHeight = height;
+            drawWidth = height * videoAspect;
+            offsetX = (width - drawWidth) / 2;
+            offsetY = 0;
+        } else {
+            drawWidth = width;
+            drawHeight = width / videoAspect;
+            offsetX = 0;
+            offsetY = (height - drawHeight) / 2;
         }
+
+        canvasCtx.drawImage(video, offsetX, offsetY, drawWidth, drawHeight);
     }
-    // Always schedule next frame, even if paused (keeps loop alive for resume)
+    // Always schedule next frame
     animationFrameId = requestAnimationFrame(drawVideo);
 }
 
@@ -219,13 +216,13 @@ function resizeCanvas() {
 }
 window.addEventListener('resize', resizeCanvas);
 
-// Streamed generation (with resize, FormData, retry, timeout; removed stub as if(isGenerating) handles pause)
+// Streamed generation (with resize, FormData, retry, timeout)
 async function generateCaption(retryCount = 0) {
     const now = Date.now();
     if (isGenerating || (now - lastGenerationTime < DEBOUNCE_MS)) return;
     isGenerating = true;
     lastGenerationTime = now;
-    const resizedDataUrl = resizeImageForUpload(canvasElement);  // Key: smaller image
+    const resizedDataUrl = resizeImageForUpload(canvasElement);  // Snapshot at start
     const promptToUse = getPromptToUse();
     captionDiv.textContent = 'Generating...';  // Better UX
     sendBtn.disabled = true;
@@ -239,7 +236,7 @@ async function generateCaption(retryCount = 0) {
         micBtn.classList.add('processing');
     }
 
-    // No need to stub drawVideo; if(isGenerating) in drawVideo skips drawing but keeps RAF loop alive
+    // Live feed continues independently via drawVideo loop
 
     try {
         const formData = new FormData();
@@ -282,7 +279,6 @@ async function generateCaption(retryCount = 0) {
     } finally {
         isGenerating = false;
         sendBtn.disabled = false;
-        // No need to restore drawVideo; loop is always alive
         if (wasVoiceActive) {
             promptInput.value = '';
             isPausedForGeneration = false;