Manual illustrations: six nano-banana images, web-sized JPEGs

Generated with tools/gen_manual_art.py (gemini-3-pro / 3.1-flash /
2.5-flash), consistent picture-book style, no text in images, ~480KB total.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

.gitignore

@@ -1,2 +1,3 @@
 .DS_Store
 __pycache__/
+googlers.txt

manual.html

@@ -33,7 +33,7 @@
     <div class="sub">Bilateral stimulation for therapy — on whatever screen you already have.</div>
 
     <div class="step">
-      <img src="img/manual/01-cover.png" alt="" onerror="this.style.display='none'">
+      <img src="img/manual/01-cover.jpg" alt="" onerror="this.style.display='none'">
       <h2>What is this?</h2>
       <p>A gentle moving dot (and optional sound) that a patient follows with
       their eyes while you guide the session. It runs in any web browser —
@@ -41,7 +41,7 @@
     </div>
 
     <div class="step">
-      <img src="img/manual/02-open.png" alt="" onerror="this.style.display='none'">
+      <img src="img/manual/02-open.jpg" alt="" onerror="this.style.display='none'">
       <h2><span class="n">1</span>Open it</h2>
       <p>Open the site on your computer. You'll see one big green
       <strong>START</strong> button and a few large sliders. That's the whole
@@ -49,7 +49,7 @@
     </div>
 
     <div class="step">
-      <img src="img/manual/03-link.png" alt="" onerror="this.style.display='none'">
+      <img src="img/manual/03-link.jpg" alt="" onerror="this.style.display='none'">
       <h2><span class="n">2</span>Give the patient a screen</h2>
       <p>Easiest: tap <strong>Open patient screen</strong> and turn that window
       to face them. Or tap <strong>Copy link</strong> and open it on their
@@ -57,7 +57,7 @@
     </div>
 
     <div class="step">
-      <img src="img/manual/04-follow.png" alt="" onerror="this.style.display='none'">
+      <img src="img/manual/04-follow.jpg" alt="" onerror="this.style.display='none'">
       <h2><span class="n">3</span>Press start</h2>
       <p>The dot glides left and right. The patient follows it with their
       eyes while you lead the session. Sound can tick from side to side too
@@ -65,7 +65,7 @@
     </div>
 
     <div class="step">
-      <img src="img/manual/05-adjust.png" alt="" onerror="this.style.display='none'">
+      <img src="img/manual/05-adjust.jpg" alt="" onerror="this.style.display='none'">
       <h2><span class="n">4</span>Make it comfortable</h2>
       <p>Slide between turtle and rabbit until the pace feels right. Change
       the size, the color, the sound — the patient's screen updates
@@ -73,7 +73,7 @@
     </div>
 
     <div class="step safety">
-      <img src="img/manual/06-care.png" alt="" onerror="this.style.display='none'">
+      <img src="img/manual/06-care.jpg" alt="" onerror="this.style.display='none'">
       <h2>⚠️ Go gently</h2>
       <p>Ask about seizure history before showing moving or pulsing visuals.
       Bilateral stimulation can bring up strong feelings — this tool assists

tools/gen_manual_art.py

@@ -11,8 +11,8 @@ KEY = os.environ.get("GEMINI_API_KEY")
 if not KEY:
     sys.exit('set GEMINI_API_KEY in the environment first')
 
-MODELS = ['gemini-2.5-flash-image', 'gemini-2.5-flash-image-preview',
+MODELS = ['gemini-3-pro-image', 'gemini-3.1-flash-image',
-          'gemini-2.0-flash-preview-image-generation']
+          'gemini-2.5-flash-image']
 
 STYLE = ("Flat vector cartoon illustration in a warm, friendly picture-book style. "
          "Soft rounded shapes, thick clean outlines, muted palette of deep teal, warm coral, "
@@ -52,15 +52,17 @@ def generate(prompt):
                '%s:generateContent?key=%s' % (model, KEY))
         req = urllib.request.Request(url, data=body,
                                      headers={'Content-Type': 'application/json'})
+        for attempt in (1, 2):
             try:
-            with urllib.request.urlopen(req, timeout=120) as r:
+                with urllib.request.urlopen(req, timeout=300) as r:
                     data = json.load(r)
                 for part in data['candidates'][0]['content']['parts']:
                     if 'inlineData' in part:
                         return base64.b64decode(part['inlineData']['data']), model
                 last = 'no image part in response from %s' % model
+                break  # empty response: don't retry same model, fall through
             except Exception as e:
-            last = '%s: %s' % (model, e)
+                last = '%s (try %d): %s' % (model, attempt, e)
     raise RuntimeError(last)