Encode the splatter bed too, and give the opponent his racket back

zombie-mode: 5.9 MB of 24-bit WAV down to 287 KB. Its master is 982597 frames
at 48 kHz, so loopEnd is the un-round 20.47077 — the decoder hands back
20.4787, and those 7.9 ms of AAC padding would tick once a loop. Nothing in
assets/audio/ is a multi-megabyte WAV any more.

o_run1/o_run2 had no racket in either hand. The cause was the batch prompt:
opponent_run never said "racket", and its reference sheet was built from the
racket-less frames it was replacing, so there was nothing to copy one from. It
is a ref_only batch off o_idle1 now — same man, front-facing, racket in his
right hand — with the subject spelling out that he still carries it mid-stride.
The stride reads as well as the old one did.

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

assets/_regen_batches.json

@@ -145,8 +145,12 @@
    "o_run1",
    "o_run2"
   ],
-  "subject": "Two frames of the same tennis opponent in a RED shirt and dark shorts running across the court towards the camera, two stages of one stride, drawn as coarse low-resolution pixel art.",
-  "anchor": 0
+  "ref_only": [
+   "o_idle1"
+  ],
+  "anchor": 0,
+  "check": "same",
+  "subject": "The reference is ONE tennis opponent in a RED polo shirt, dark grey shorts and a white headband, seen from the FRONT, gripping a tennis racket in his RIGHT hand. Draw TWO frames of that SAME man, same size, same clothes, same face, same front-facing view, RUNNING across the court towards the camera as two stages of one stride: the first with his left leg forward and his right leg trailing, the second with the legs reversed. In BOTH frames he is STILL HOLDING THE TENNIS RACKET in his right hand, the same racket as the reference \u2014 head, strings, throat and grip all drawn, carried out to his side at about waist height as a runner carries it. His other arm swings free. The racket must be clearly visible and fully inside the canvas in both frames; neither hand is ever empty."
  },
  "ump_boy": {
   "sprites": [