diff --git a/training-data/schema.json b/training-data/schema.json
new file mode 100644
index 00000000..73c9ad31
--- /dev/null
+++ b/training-data/schema.json
@@ -0,0 +1,84 @@
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "Scene Description Training Entry",
+  "description": "Schema for lyrics-to-visual-scene description training data entries.",
+  "type": "object",
+  "required": ["song", "artist", "beat", "timestamp", "lyric_line", "scene"],
+  "properties": {
+    "song": {
+      "type": "string",
+      "minLength": 1,
+      "description": "Song title"
+    },
+    "artist": {
+      "type": "string",
+      "minLength": 1,
+      "description": "Artist name"
+    },
+    "mood_arc": {
+      "type": "string",
+      "description": "Overall mood progression of the song (optional)"
+    },
+    "beat": {
+      "type": "integer",
+      "minimum": 1,
+      "description": "Beat number within the song (1-indexed)"
+    },
+    "timestamp": {
+      "type": "string",
+      "pattern": "^[0-9]+:[0-5][0-9]$",
+      "description": "Timestamp in M:SS or MM:SS format"
+    },
+    "duration": {
+      "type": "string",
+      "description": "Duration of the beat (e.g. '30s', '15s')"
+    },
+    "duration_seconds": {
+      "type": "integer",
+      "minimum": 1,
+      "description": "Duration in seconds (integer alternative)"
+    },
+    "lyric_line": {
+      "type": "string",
+      "minLength": 1,
+      "description": "The lyric line for this beat"
+    },
+    "scene": {
+      "type": "object",
+      "required": ["mood", "colors", "composition", "description"],
+      "properties": {
+        "mood": {
+          "type": "string",
+          "minLength": 1,
+          "description": "Emotional mood of the scene"
+        },
+        "colors": {
+          "type": "array",
+          "items": { "type": "string", "minLength": 1 },
+          "minItems": 1,
+          "description": "Color palette for the scene"
+        },
+        "composition": {
+          "type": "string",
+          "minLength": 1,
+          "description": "Shot composition (e.g. 'wide shot', 'close-up', 'low angle')"
+        },
+        "camera": {
+          "type": "string",
+          "description": "Camera movement (e.g. 'static', 'slow zoom', 'tracking')"
+        },
+        "camera_movement": {
+          "type": "string",
+          "description": "Alternative field name for camera movement"
+        },
+        "description": {
+          "type": "string",
+          "minLength": 10,
+          "description": "Full scene description text"
+        }
+      },
+      "additionalProperties": true
+    }
+  },
+  "additionalProperties": true
+}