diff --git a/backend/routers/voice.py b/backend/routers/voice.py
index a8b6005..63eeed6 100644
--- a/backend/routers/voice.py
+++ b/backend/routers/voice.py
@@ -38,6 +38,7 @@ class ChatIn(BaseModel):
     session_key: str = ""           # optional: Langzeit-Memory-Scope
     system: str = ""                # optionaler ephemerer System-Prompt (z.B. „antworte knapp/gesprochen")
     model: str = ""
+    image: str = ""                 # optionale Bildschirm-Sicht: Bild als data:-URL (Lucys „Augen")
 
 
 @router.get("/voice/health")
@@ -139,7 +140,15 @@ async def voice_chat(body: ChatIn) -> StreamingResponse:
     messages = []
     if body.system:
         messages.append({"role": "system", "content": body.system})
-    messages.append({"role": "user", "content": body.text})
+    if body.image:
+        # Bildschirm-Sicht: User-Message als multimodal (Text + Bild). Das Agent-Modell (fast/Qwen3.6
+        # mit mmproj) bzw. ein Vision-Modell verarbeitet das Bild via OpenAI-image_url.
+        messages.append({"role": "user", "content": [
+            {"type": "text", "text": body.text},
+            {"type": "image_url", "image_url": {"url": body.image}},
+        ]})
+    else:
+        messages.append({"role": "user", "content": body.text})
     payload = {"model": body.model or HERMES_API_MODEL, "messages": messages, "stream": True}
     headers = {
         "Authorization": f"Bearer {HERMES_API_KEY}",