A golden retriever sprinting through shallow surf on a beach at golde...
The prompt
A golden retriever sprinting through shallow surf on a beach at golden hour, slow motion, water droplets catching the sun, fur rippling, photorealistic wildlife cinematography, telephoto lens, shallow depth of field
What the agent wrote
A golden retriever sprints through ankle-deep shallow surf along a wide sandy beach at golden hour, captured in crisp slow motion. The camera tracks alongside the dog at running speed, telephoto lens compressing the warm hazy background into soft bokeh with shallow depth of field. Sea water sprays and droplets scatter upward, each bead catching the low amber sun. The dog's wet golden fur ripples and bounces with every powerful stride, ears flapping, paws kicking up bright splashes. Backlit rim light glows along its coat. Photorealistic wildlife cinematography, warm dreamy mood, continuous unbroken shot.
Why this framing: A lateral tracking shot of a running subject across a wide beach suits landscape 16:9, giving room for the dog's horizontal motion and the telephoto background compression.
The eval receipt
Every Take clip ships with the measurements that admitted it. Deterministic gates and CV lanes run first; the VLM judge only rules on what those lanes cannot measure, and it must justify each level before naming it. How the cascade works.
L0 · Deterministic gates
| Gate | Result | Reading | Check |
|---|---|---|---|
| decodes | pass | h264 | ffprobe parsed a video stream |
| duration | pass | 5.04s | expected 5s ± 1s |
| resolution | pass | 1280x720 | expected height 720 |
| framerate | pass | 24.00 fps | sane range 12-60 |
| not black | pass | mean luma 96.2 | 0/121 frames under 12.0 |
| not frozen | pass | mean frame diff 6.523 | static-cheat detector, threshold 0.35 |
L1 · CV lanes
mean abs luma diff between consecutive frames
How much each frame differs from the next, checked on every frame. A high average usually just means a busy scene; sudden spikes far above the average mean strobing.
Farneback flow magnitude, motion energy per frame pair
How far pixels move between frames: one number for how much is actually happening. Under 0.3 is basically a still image; 2 to 8 is normal motion.
cosine of CLIP ViT-B/32 prompt and frame embeddings
How well the frames match what you typed, scored by CLIP. 0.30 and up is well on-prompt; below 0.24 the model probably wandered off.
DINOv2 cosine of each sampled frame against frame 0
Whether the subject stays the same subject, with every frame compared against the first. Watch the min: a single frame below 0.5 means identity broke.
These bands come from a 14-clip calibration set we ran before launch. See the full thresholds and the clips that set them.
L2 · Judge verdict
A polished, photorealistic golden-hour clip of a golden retriever bounding through shallow surf, with convincing splashes, warm rim light, and a soft bokeh background that closely matches the prompt. Motion is steady and natural, the dog's identity holds throughout (the moderate DINO drift just reflects it charging toward the lens), and water physics look believable. The only minor mismatch is that the shot reads more head-on than a strict side-tracking move.
What the judge saw
The timestamped contact sheet, exactly as handed to the judge. ffmpeg pulled eight evenly spaced frames from the clip, stamped each with its timestamp, and tiled them into this one grid; it is the only image the judge reads. Why a grid beats a video.
