Back to all instances

Ireland

Benchmark instance 040
Input static SVG
Input static SVG for Ireland
Source Lottie animation
Motion edit prompt

Frame 0 already matches the target: use the provided SVG as-is, with the existing clipped vertical letter stacks; no elements need to be moved, scaled, or hidden before animation begins. Keep each letter’s horizontal position, scale, and orientation fixed. For every letter in “IRELAND,” use its existing vertical stack of identical copies and its existing clip to the original word-height viewing area so only one letter copy is normally visible. After a short hold, animate the stacks sliding upward in a left-to-right cascade: I moves first, followed by R, E, L, A, N, and D with small even delays. Each stack moves upward by one full copy spacing with smooth eased motion, pauses, then repeats the same upward jump-and-hold cycle two more times, for three upward slides total. During each upward slide, let the outgoing letter be clipped off at the top while the next identical copy enters from the bottom, creating temporary split or cropped letter fragments inside the letter’s column. After the final cascade, all letters settle into full alignment on the baseline and return to where they sit in the provided SVG, spelling “IRELAND,” using lower copies in the same existing stacks.

Example model outputs

Human-approved

GLM-5.2

Reasoning effort: high

Verification report: 6 passed · 0 failed
Result Check
PASS ver_fixed_letter_artwork

Identical original glyphs translate vertically without artwork edits

Measured values
{
  "maximum_outline_error": 0.0006000000000767614
}
PASS ver_left_to_right_even_cascades

Evenly delayed I-to-D upward cascades

Measured values
{
  "sampled_delays": [
    [
      4.0,
      4.0,
      4.0,
      5.0,
      4.0,
      4.0
    ],
    [
      4.0,
      5.0,
      4.0,
      4.0,
      4.0,
      4.0
    ],
    [
      4.0,
      4.0,
      4.0,
      5.0,
      4.0,
      4.0
    ]
  ]
}
PASS ver_original_endpoint_poses

Original full, opaque purple word at both endpoints

Measured values
{
  "maximum_pose_error": 0.006100000000003547
}
PASS ver_original_window_clipped_handoffs

Original clipping shows top-out/bottom-in copy handoffs

Measured values
{
  "cropped_phases": {
    "A": [
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ]
    ],
    "D": [
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ]
    ],
    "E": [
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ]
    ],
    "I": [
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ]
    ],
    "L": [
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ]
    ],
    "N": [
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ],
      [
        6,
        6,
        2
      ]
    ],
    "R": [
      [
        6,
        7,
        3
      ],
      [
        6,
        6,
        2
      ],
      [
        6,
        7,
        3
      ]
    ]
  },
  "paired_fragment_samples": 44
}
PASS ver_smooth_eased_slides

Resolved eased motion, with no particular curve or duration imposed

Measured values
{
  "minimum_position_nonlinearity": 46.6891
}
PASS ver_three_upward_jump_hold_cycles

Three upward one-copy jump-and-hold cycles per stack

Measured values
{
  "copies": {
    "A": [
      1.000003429904191,
      1.000007622009314,
      1.0000034299041913
    ],
    "D": [
      1.000007622009314,
      1.000003429904191,
      1.0000076220093141
    ],
    "E": [
      1.000007622009314,
      1.0000072409088483,
      1.0000076220093141
    ],
    "I": [
      1.000007622009314,
      1.000007622009314,
      1.000007622009314
    ],
    "L": [
      1.0000072409088483,
      1.000007622009314,
      1.0000072409088485
    ],
    "N": [
      1.000007622009314,
      1.000007622009314,
      1.0000076220093141
    ],
    "R": [
      1.000007622009314,
      1.000003429904191,
      1.0000076220093141
    ]
  }
}

Human-rejected

GPT-5.6 Terra

Reasoning effort: high

Verification report: 4 passed · 2 failed
Result Check
PASS ver_fixed_letter_artwork

Identical original glyphs translate vertically without artwork edits

Measured values
{
  "maximum_outline_error": 0.0006000000000767614
}
PASS ver_left_to_right_even_cascades

Evenly delayed I-to-D upward cascades

Measured values
{
  "sampled_delays": [
    [
      3.0,
      3.0,
      3.0,
      3.0,
      3.0,
      3.0
    ],
    [
      3.0,
      3.0,
      3.0,
      3.0,
      3.0,
      3.0
    ],
    [
      3.0,
      3.0,
      3.0,
      3.0,
      3.0,
      3.0
    ]
  ]
}
FAIL ver_original_endpoint_poses

Original full, opaque purple word at both endpoints

Measured values
{
  "frame": 0,
  "position_error": 89.19599999999998,
  "visible_fraction": 0.5156
}
FAIL ver_original_window_clipped_handoffs

Original clipping shows top-out/bottom-in copy handoffs

Measured values
{
  "incoming_crops": 0,
  "outgoing_crops": 0
}
PASS ver_smooth_eased_slides

Resolved eased motion, with no particular curve or duration imposed

Measured values
{
  "minimum_position_nonlinearity": 48.3655
}
PASS ver_three_upward_jump_hold_cycles

Three upward one-copy jump-and-hold cycles per stack

Measured values
{
  "copies": {
    "A": [
      1.0,
      0.9999999999999996,
      1.0000000000000004
    ],
    "D": [
      1.0,
      0.9999999999999996,
      1.0000000000000004
    ],
    "E": [
      1.0,
      0.9999999999999996,
      1.0000000000000004
    ],
    "I": [
      1.0,
      0.9999999999999996,
      1.0000000000000004
    ],
    "L": [
      1.0,
      0.9999999999999996,
      1.0000000000000004
    ],
    "N": [
      1.0,
      0.9999999999999996,
      1.0000000000000004
    ],
    "R": [
      1.0,
      0.9999999999999996,
      1.0000000000000004
    ]
  }
}