← Back to EgoPlay

Full Results

Page 010 / 010 · 15 of 150 samples (sorted by id)

Each unit shows the 4 event-trigger modes for one Ego4D source clip. Inside each mode row: source, GT target, and the 4 method outputs, each annotated with its own VLM scores (Edit quality, Visual quality, BG consistency, and Total). The red border highlights the event window; use "Go to event start" to jump a video to its trigger frame.

Videos on this page are compressed for the web. Download the full benchmark ↓ — all videos, method outputs, and annotations (JSON + CSV).

f0351bbefb87123f7fd3ef8b6daa65c9_event4_p1

GT edit prompt: Add a tiny mossy goblin with button eyes peeking from under the black napkin, holding a miniature spoon made from acorn caps.

Single Positive

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 2.2Visual 2.2BG 2.2Total 6.8

Event prompt: Once you detect scooping the ice cream with the biscuits in the bowl with the spoon, add a tiny mossy goblin with button eyes peeking from under the black napkin, holding a miniature spoon made from acorn caps.

Before event: 20 frames (0.00s – 1.25s)  ·  Event: 42 frames (1.25s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 2.0Visual 2.0BG 2.0Total 6.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 2.0Visual 2.0BG 2.0Total 6.0

EgoPlay-Causal

Edit 2.0Visual 2.0BG 2.0Total 6.0

Event prompt: When I lift the glass of water to my lips and take a sip, add a tiny mossy goblin with button eyes peeking from under the black napkin, holding a miniature spoon made from acorn caps.

Before event: 20 frames (0.00s – 1.25s)  ·  Event: 42 frames (1.25s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 0.8Visual 0.8BG 0.8Total 2.2

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 2.2Visual 2.2BG 2.2Total 6.8

EgoPlay-Causal

Edit 1.5Visual 1.5BG 1.5Total 4.5

Event prompt: Right after I push the bowl slightly to the left edge of the table, remove the glass of water from the table, leaving no trace behind. Once you detect scooping the ice cream with the biscuits in the bowl with the spoon, add a tiny mossy goblin with button eyes peeking from under the black napkin, holding a miniature spoon made from acorn caps. As soon as I press the phone with my left hand to unlock it, change the weather outside the window to a steady downpour of rain with droplets streaking the glass. After I place the empty snack packet on the box behind the bowl, shift the lighting to warm candlelight, casting soft glows and deep shadows across the table and hands.

Before event: 20 frames (0.00s – 1.25s)  ·  Event: 42 frames (1.25s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 2.0Visual 2.0BG 2.0Total 6.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 1.0Visual 1.0BG 1.0Total 3.0

EgoPlay-Causal

Edit 2.0Visual 2.0BG 2.0Total 6.0

Event prompt: Right after I push the bowl slightly to the left edge of the table, remove the glass of water from the table, leaving no trace behind. When I lift the glass of water to my lips and take a sip, add a tiny mossy goblin with button eyes peeking from under the black napkin, holding a miniature spoon made from acorn caps. As soon as I press the phone with my left hand to unlock it, change the weather outside the window to a steady downpour of rain with droplets streaking the glass. After I place the empty snack packet on the box behind the bowl, shift the lighting to warm candlelight, casting soft glows and deep shadows across the table and hands.

Before event: 20 frames (0.00s – 1.25s)  ·  Event: 42 frames (1.25s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

f2d9f9ce9cde07ef38cdd9de6f4c2d4e_event0_p1

GT edit prompt: Recolor the red cover of the open book to deep indigo while preserving its matte texture and all printed text.

Single Positive

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once you detect me moving my left hand under the book, recolor the red cover of the open book to deep indigo while preserving its matte texture and all printed text.

Before event: 2 frames (0.00s – 0.12s)  ·  Event: 48 frames (0.12s – 3.12s)  ·  After event: 31 frames (3.12s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I lift the book off my lap to place it on the floor, recolor the red cover of the open book to deep indigo while preserving its matte texture and all printed text.

Before event: 2 frames (0.00s – 0.12s)  ·  Event: 48 frames (0.12s – 3.12s)  ·  After event: 31 frames (3.12s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 2.2Visual 2.2BG 2.2Total 6.8

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once you detect me moving my left hand under the book, recolor the red cover of the open book to deep indigo while preserving its matte texture and all printed text. Immediately after I wipe my fingers on the cloth next to me, replace the background with a calm indoor study, featuring wooden shelves lined with books and a small potted plant near the window. As soon as I adjust the sleeve of my right arm, transform the image into a watercolor painting style, preserving all details of the book, hands, and clothing with soft edges and blended colors. Once I shift the green glass on the ground beside the book, add a small ceramic teacup with steam rising beside the book on the table, oriented to the left side of the frame.

Before event: 2 frames (0.00s – 0.12s)  ·  Event: 48 frames (0.12s – 3.12s)  ·  After event: 31 frames (3.12s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I lift the book off my lap to place it on the floor, recolor the red cover of the open book to deep indigo while preserving its matte texture and all printed text. Immediately after I wipe my fingers on the cloth next to me, replace the background with a calm indoor study, featuring wooden shelves lined with books and a small potted plant near the window. As soon as I adjust the sleeve of my right arm, transform the image into a watercolor painting style, preserving all details of the book, hands, and clothing with soft edges and blended colors. Once I shift the green glass on the ground beside the book, add a small ceramic teacup with steam rising beside the book on the table, oriented to the left side of the frame.

Before event: 2 frames (0.00s – 0.12s)  ·  Event: 48 frames (0.12s – 3.12s)  ·  After event: 31 frames (3.12s – 5.06s)

f348b04d8c7bb59b926b5d6ff054d3c1_event0_p2

GT edit prompt: Swap the curved wooden object in your left hand for a matte black ceramic sculpture shaped like a wave, retaining the same grip and orientation.

Single Positive

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 1.0Visual 1.0BG 1.0Total 3.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: When I spray the wood with spray paint, swap the curved wooden object in my left hand for a matte black ceramic sculpture shaped like a wave, retaining the same grip and orientation.

Before event: 0 frames (0.00s – 0.00s)  ·  Event: 43 frames (0.00s – 2.69s)  ·  After event: 38 frames (2.69s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: When I wipe the paint can lid with a cloth, swap the curved wooden object in your left hand for a matte black ceramic sculpture shaped like a wave, retaining the same grip and orientation.

Before event: 0 frames (0.00s – 0.00s)  ·  Event: 43 frames (0.00s – 2.69s)  ·  After event: 38 frames (2.69s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 1.0Visual 1.0BG 1.0Total 3.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Immediately after I pour paint from the can onto the cardboard, change the scene to midday with bright overhead sunlight casting short, defined shadows. When I spin a wooden piece on the ground with my foot, shift the lighting to a warm golden glow typical of late afternoon, with soft long shadows across the floor. When I spray the wood with spray paint, swap the curved wooden object in my left hand for a matte black ceramic sculpture shaped like a wave, retaining the same grip and orientation.

Before event: 0 frames (0.00s – 0.00s)  ·  Event: 43 frames (0.00s – 2.69s)  ·  After event: 38 frames (2.69s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Immediately after I pour paint from the can onto the cardboard, change the scene to midday with bright overhead sunlight casting short, defined shadows. When I spin a wooden piece on the ground with my foot, shift the lighting to a warm golden glow typical of late afternoon, with soft long shadows across the floor. When I wipe the paint can lid with a cloth, swap the curved wooden object in your left hand for a matte black ceramic sculpture shaped like a wave, retaining the same grip and orientation.

Before event: 0 frames (0.00s – 0.00s)  ·  Event: 43 frames (0.00s – 2.69s)  ·  After event: 38 frames (2.69s – 5.06s)

f3d9a98382207e82367b232303c05674_event1_p3

GT edit prompt: Remove the pink cutting board from the sink ledge and restore the metal surface beneath it seamlessly.

Single Positive

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I have finished closing the cabinet, remove the pink cutting board from the sink ledge and restore the metal surface beneath it seamlessly.

Before event: 5 frames (0.00s – 0.31s)  ·  Event: 57 frames (0.31s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: As soon as I pick up the plastic cup from the sink, remove the pink cutting board from the sink ledge and restore the metal surface beneath it seamlessly.

Before event: 5 frames (0.00s – 0.31s)  ·  Event: 57 frames (0.31s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: As soon as I wipe the countertop near the stove with a cloth, change the texture of the frying pan’s surface to a hammered metal pattern, while preserving its red color and position. Once I have finished closing the cabinet, remove the pink cutting board from the sink ledge and restore the metal surface beneath it seamlessly. Once I grab the packet of spices from the shelf above the sink, add a thin wisp of steam rising from the frying pan, settling gently above the food.

Before event: 5 frames (0.00s – 0.31s)  ·  Event: 57 frames (0.31s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: As soon as I wipe the countertop near the stove with a cloth, change the texture of the frying pan’s surface to a hammered metal pattern, while preserving its red color and position. As soon as I pick up the plastic cup from the sink, remove the pink cutting board from the sink ledge and restore the metal surface beneath it seamlessly. Once I grab the packet of spices from the shelf above the sink, add a thin wisp of steam rising from the frying pan, settling gently above the food.

Before event: 5 frames (0.00s – 0.31s)  ·  Event: 57 frames (0.31s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

f540a27edc0ae08f34bd1664650bc043_event5_p3

GT edit prompt: Add a polished steel wristband with rivet detailing to the left wrist, catching the light as the hand holds the brake pad.

Single Positive

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: After you remove the brake pad from the pack with your right hand, add a polished steel wristband with rivet detailing to your left wrist, catching the light as you hold the pad.

Before event: 37 frames (0.00s – 2.31s)  ·  Event: 24 frames (2.31s – 3.81s)  ·  After event: 20 frames (3.81s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Immediately after I adjust the position of the box on the workbench with my left hand, add a polished steel wristband with rivet detailing to the left wrist, catching the light as the hand holds the brake pad.

Before event: 37 frames (0.00s – 2.31s)  ·  Event: 24 frames (2.31s – 3.81s)  ·  After event: 20 frames (3.81s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: After I pick up the metal cylinder from the left side of the workbench, add a small screwdriver to the right side of the workbench, placed beside the drill, with the handle resting on the surface and the tip pointed upward. After I push the plastic bag toward the center of the workbench, transform the person into a cyborg with metallic arms and a wrist-mounted display, keeping the same posture and workbench setup. After you remove the brake pad from the pack with your right hand, add a polished steel wristband with rivet detailing to your left wrist, catching the light as you hold the pad.

Before event: 37 frames (0.00s – 2.31s)  ·  Event: 24 frames (2.31s – 3.81s)  ·  After event: 20 frames (3.81s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: After I pick up the metal cylinder from the left side of the workbench, add a small screwdriver to the right side of the workbench, placed beside the drill, with the handle resting on the surface and the tip pointed upward. After I push the plastic bag toward the center of the workbench, transform the person into a cyborg with metallic arms and a wrist-mounted display, keeping the same posture and workbench setup. Immediately after I adjust the position of the box on the workbench with my left hand, add a polished steel wristband with rivet detailing to the left wrist, catching the light as the hand holds the brake pad.

Before event: 37 frames (0.00s – 2.31s)  ·  Event: 24 frames (2.31s – 3.81s)  ·  After event: 20 frames (3.81s – 5.06s)

f6aa7cabac2072e14a1a38b61d9b5724_event0_p0

GT edit prompt: Switch the ripped jeans and sandals to tailored khaki pants and black work boots, maintaining the same stance and lighting.

Single Positive

Source

GT Target

EgoEdit

Edit 2.2Visual 2.2BG 2.2Total 6.8

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 2.2Visual 2.2BG 2.2Total 6.8

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Do switching my ripped jeans and sandals to tailored khaki pants and black work boots, maintaining the same stance and lighting, once I move my leg is done.

Before event: 1 frames (0.00s – 0.06s)  ·  Event: 27 frames (0.06s – 1.75s)  ·  After event: 53 frames (1.75s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I lift the metal rack by its top bar, switch the ripped jeans and sandals to tailored khaki pants and black work boots, maintaining the same stance and lighting.

Before event: 1 frames (0.00s – 0.06s)  ·  Event: 27 frames (0.06s – 1.75s)  ·  After event: 53 frames (1.75s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 2.2Visual 2.2BG 2.2Total 6.8

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 2.2Visual 2.2BG 2.2Total 6.8

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Do switching my ripped jeans and sandals to tailored khaki pants and black work boots, maintaining the same stance and lighting, once I move my leg is done. Immediately after I pick up the second angle grinder from the ground, change the metal grating to a wooden slatted bench, maintaining the same size, orientation, and placement.

Before event: 1 frames (0.00s – 0.06s)  ·  Event: 27 frames (0.06s – 1.75s)  ·  After event: 53 frames (1.75s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 1.0Visual 1.0BG 1.0Total 3.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I lift the metal rack by its top bar, switch the ripped jeans and sandals to tailored khaki pants and black work boots, maintaining the same stance and lighting. Immediately after I pick up the second angle grinder from the ground, change the metal grating to a wooden slatted bench, maintaining the same size, orientation, and placement.

Before event: 1 frames (0.00s – 0.06s)  ·  Event: 27 frames (0.06s – 1.75s)  ·  After event: 53 frames (1.75s – 5.06s)

f6fdba5011cafda1867980efba398e1b_event2_p2

GT edit prompt: Swap the white plastic container in your left hand with a small copper tea kettle, same size and grip orientation, with realistic metal highlights.

Single Positive

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I pick a container from the sink occurs, swap the white plastic container in my left hand with a small copper tea kettle, same size and grip orientation, with realistic metal highlights.

Before event: 26 frames (0.00s – 1.62s)  ·  Event: 49 frames (1.62s – 4.69s)  ·  After event: 6 frames (4.69s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 2.3BG 3.0Total 8.3

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 2.3BG 3.0Total 8.3

EgoPlay-Causal

Edit 3.0Visual 2.3BG 3.0Total 8.3

Event prompt: Right after I wipe the countertop near the knife block with a cloth, swap the white plastic container in your left hand with a small copper tea kettle, same size and grip orientation, with realistic metal highlights.

Before event: 26 frames (0.00s – 1.62s)  ·  Event: 49 frames (1.62s – 4.69s)  ·  After event: 6 frames (4.69s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I pick a container from the sink occurs, swap the white plastic container in my left hand with a small copper tea kettle, same size and grip orientation, with realistic metal highlights. Right after I lift the kettle from the stovetop to pour water, after the person places the bowl in the dishwasher, add a single shiny red pepper lying on the countertop near the sink, positioned to the left of the dish rack with realistic shadows.

Before event: 26 frames (0.00s – 1.62s)  ·  Event: 49 frames (1.62s – 4.69s)  ·  After event: 6 frames (4.69s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 2.3BG 3.0Total 8.3

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Right after I wipe the countertop near the knife block with a cloth, swap the white plastic container in your left hand with a small copper tea kettle, same size and grip orientation, with realistic metal highlights. Right after I lift the kettle from the stovetop to pour water, after the person places the bowl in the dishwasher, add a single shiny red pepper lying on the countertop near the sink, positioned to the left of the dish rack with realistic shadows.

Before event: 26 frames (0.00s – 1.62s)  ·  Event: 49 frames (1.62s – 4.69s)  ·  After event: 6 frames (4.69s – 5.06s)

f71a9f93d94da48c0ae8e192c0483c62_event1_p1

GT edit prompt: Transform the tablet’s casing into weathered brass with subtle patina and fingerprint smudges, preserving its exact shape and position.

Single Positive

Source

GT Target

EgoEdit

Edit 0.0Visual 0.0BG 1.5Total 1.5

VLM-guided

Edit 1.5Visual 1.5BG 3.0Total 6.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: When I write on the tablet, transform the tablet’s casing into weathered brass with subtle patina and fingerprint smudges, preserving its exact shape and position.

Before event: 22 frames (0.00s – 1.38s)  ·  Event: 31 frames (1.38s – 3.31s)  ·  After event: 28 frames (3.31s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Right after I adjust the position of the monitor by pushing it slightly to the left, transform the tablet’s casing into weathered brass with subtle patina and fingerprint smudges, preserving its exact shape and position.

Before event: 22 frames (0.00s – 1.38s)  ·  Event: 31 frames (1.38s – 3.31s)  ·  After event: 28 frames (3.31s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 3.0Total 6.0

VLM-guided

Edit 1.5Visual 1.5BG 3.0Total 6.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 1.5Visual 1.5BG 1.5Total 4.5

Event prompt: Once I pick up the headphones from the desk and place them on my neck, change the color of the tablet screen display to vivid purple while keeping the interface elements and lines unchanged. When I write on the tablet, transform the tablet’s casing into weathered brass with subtle patina and fingerprint smudges, preserving its exact shape and position. After I slide the keyboard closer to the edge of the desk with my left hand, replace the tablet with a closed notebook of the same size and orientation, keeping the hand position and surrounding objects unchanged.

Before event: 22 frames (0.00s – 1.38s)  ·  Event: 31 frames (1.38s – 3.31s)  ·  After event: 28 frames (3.31s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 2.0Visual 2.0BG 2.0Total 6.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I pick up the headphones from the desk and place them on my neck, change the color of the tablet screen display to vivid purple while keeping the interface elements and lines unchanged. Right after I adjust the position of the monitor by pushing it slightly to the left, transform the tablet’s casing into weathered brass with subtle patina and fingerprint smudges, preserving its exact shape and position. After I slide the keyboard closer to the edge of the desk with my left hand, replace the tablet with a closed notebook of the same size and orientation, keeping the hand position and surrounding objects unchanged.

Before event: 22 frames (0.00s – 1.38s)  ·  Event: 31 frames (1.38s – 3.31s)  ·  After event: 28 frames (3.31s – 5.06s)

f82c7ced8cf8be2e6d215ea20e79c855_event4_p1

GT edit prompt: Replace the beige plastic jerrycan in my right hand with a clay amphora sealed with wax, maintaining the same grip and tilt.

Single Positive

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: After opening the lid of the jerrycan with my left hand is completed, replace the beige plastic jerrycan in my right hand with a clay amphora sealed with wax, maintaining the same grip and tilt.

Before event: 34 frames (0.00s – 2.12s)  ·  Event: 43 frames (2.12s – 4.81s)  ·  After event: 4 frames (4.81s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: After I wipe the chain saw's engine with my sleeve, replace the beige plastic jerrycan in my right hand with a clay amphora sealed with wax, maintaining the same grip and tilt.

Before event: 34 frames (0.00s – 2.12s)  ·  Event: 43 frames (2.12s – 4.81s)  ·  After event: 4 frames (4.81s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Right after I pick up the red tape roll from the ground, after the event, add a lizard sitting on a rock in the background to the right of the scene. Right after I clean the dirt off the blade's edge with my fingers, after the event, change the material of the jerrycan to shiny aluminum with a metallic reflection. When I adjust the straps of my backpack behind me, after the event, restyle the entire image in a pixel art style with blocky colors and hard edges, preserving all object positions. After opening the lid of the jerrycan with my left hand is completed, replace the beige plastic jerrycan in my right hand with a clay amphora sealed with wax, maintaining the same grip and tilt.

Before event: 34 frames (0.00s – 2.12s)  ·  Event: 43 frames (2.12s – 4.81s)  ·  After event: 4 frames (4.81s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Right after I pick up the red tape roll from the ground, after the event, add a lizard sitting on a rock in the background to the right of the scene. Right after I clean the dirt off the blade's edge with my fingers, after the event, change the material of the jerrycan to shiny aluminum with a metallic reflection. When I adjust the straps of my backpack behind me, after the event, restyle the entire image in a pixel art style with blocky colors and hard edges, preserving all object positions. After I wipe the chain saw's engine with my sleeve, replace the beige plastic jerrycan in my right hand with a clay amphora sealed with wax, maintaining the same grip and tilt.

Before event: 34 frames (0.00s – 2.12s)  ·  Event: 43 frames (2.12s – 4.81s)  ·  After event: 4 frames (4.81s – 5.06s)

f97d098d744ae53655680069b7e16c93_event0_p0

GT edit prompt: Transform the circuit board into a pulsing network of glowing bioluminescent fungi that retain its original shape and component layout.

Single Positive

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 2.5Visual 2.5BG 3.0Total 8.0

EgoPlay

Edit 2.2Visual 2.2BG 2.2Total 6.8

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Following me touching the motherboard with my left hand, transform the circuit board into a pulsing network of glowing bioluminescent fungi that retain its original shape and component layout.

Before event: 2 frames (0.00s – 0.12s)  ·  Event: 32 frames (0.12s – 2.12s)  ·  After event: 47 frames (2.12s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 1.0Visual 1.0BG 1.0Total 3.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I pull the power supply unit out of the case with my right hand, transform the circuit board into a pulsing network of glowing bioluminescent fungi that retain its original shape and component layout.

Before event: 2 frames (0.00s – 0.12s)  ·  Event: 32 frames (0.12s – 2.12s)  ·  After event: 47 frames (2.12s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 2.2Visual 2.2BG 2.2Total 6.8

EgoPlay

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Following me touching the motherboard with my left hand, transform the circuit board into a pulsing network of glowing bioluminescent fungi that retain its original shape and component layout. Once I wipe dust off the surface of the CPU box with my left sleeve, transform the person into a cyborg with exposed mechanical limbs and wires along the arms, maintaining the same posture and position.

Before event: 2 frames (0.00s – 0.12s)  ·  Event: 32 frames (0.12s – 2.12s)  ·  After event: 47 frames (2.12s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 0.0Visual 0.0BG 0.0Total 0.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Once I pull the power supply unit out of the case with my right hand, transform the circuit board into a pulsing network of glowing bioluminescent fungi that retain its original shape and component layout. Once I wipe dust off the surface of the CPU box with my left sleeve, transform the person into a cyborg with exposed mechanical limbs and wires along the arms, maintaining the same posture and position.

Before event: 2 frames (0.00s – 0.12s)  ·  Event: 32 frames (0.12s – 2.12s)  ·  After event: 47 frames (2.12s – 5.06s)

fa4130180fb4b96f488f3117f8298851_event0_p3

GT edit prompt: Replace the black remote in my hand with a vintage wooden-handled TV dial of the same size and grip, unchanged in orientation.

Single Positive

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 0.8Visual 0.8BG 0.8Total 2.2

Event prompt: Following you lifting the remote with your left hand, replace the black remote in your hand with a vintage wooden-handled TV dial of the same size and grip, unchanged in orientation.

Before event: 44 frames (0.00s – 2.75s)  ·  Event: 18 frames (2.75s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Immediately after I adjust the TV stand's cable management with my right hand, replace the black remote in my hand with a vintage wooden-handled TV dial of the same size and grip, unchanged in orientation.

Before event: 44 frames (0.00s – 2.75s)  ·  Event: 18 frames (2.75s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 0.8Visual 0.8BG 0.8Total 2.2

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay-Causal

Edit 1.5Visual 1.5BG 1.5Total 4.5

Event prompt: Following you lifting the remote with your left hand, replace the black remote in your hand with a vintage wooden-handled TV dial of the same size and grip, unchanged in orientation. When I turn off the wall socket switch with my right index finger, transform the man on the television screen into a pirate with an eye patch and a cutlass, keeping his seat position and screen framing unchanged. After I wipe the television screen with my right sleeve, add a loaf of freshly baked bread on the wooden side table to the left of the TV, positioned at the edge near the table's corner. Once I pick up the wooden coffee table's cloth with my left hand, replace the interior room background with a bustling city street at dusk, preserving the man on the screen and all surrounding furniture in the same position and scale.

Before event: 44 frames (0.00s – 2.75s)  ·  Event: 18 frames (2.75s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 1.0Visual 1.0BG 1.0Total 3.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Immediately after I adjust the TV stand's cable management with my right hand, replace the black remote in my hand with a vintage wooden-handled TV dial of the same size and grip, unchanged in orientation. When I turn off the wall socket switch with my right index finger, transform the man on the television screen into a pirate with an eye patch and a cutlass, keeping his seat position and screen framing unchanged. After I wipe the television screen with my right sleeve, add a loaf of freshly baked bread on the wooden side table to the left of the TV, positioned at the edge near the table's corner. Once I pick up the wooden coffee table's cloth with my left hand, replace the interior room background with a bustling city street at dusk, preserving the man on the screen and all surrounding furniture in the same position and scale.

Before event: 44 frames (0.00s – 2.75s)  ·  Event: 18 frames (2.75s – 3.88s)  ·  After event: 19 frames (3.88s – 5.06s)

fae371351dff253c6b1ae5dc62326929_event0_p2

GT edit prompt: Shift the air to a fine, warm dust haze rising from the dry soil, softening distant edges while keeping my hands crisp.

Single Positive

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 1.0Visual 1.0BG 1.0Total 3.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Only after I remove the tube handle should the air shift to a fine, warm dust haze rising from the dry soil, softening distant edges while keeping my hands crisp.

Before event: 0 frames (0.00s – 0.00s)  ·  Event: 40 frames (0.00s – 2.50s)  ·  After event: 41 frames (2.50s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: As soon as I tie the loose end of the drip line to a stake in the soil, shift the air to a fine, warm dust haze rising from the dry soil, softening distant edges while keeping my hands crisp.

Before event: 0 frames (0.00s – 0.00s)  ·  Event: 40 frames (0.00s – 2.50s)  ·  After event: 41 frames (2.50s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 1.0Visual 1.0BG 1.0Total 3.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 0.7Visual 0.7BG 0.0Total 1.3

Event prompt: When I pick up the detached connector and examine it closely, replace the black plastic connector with a brass fitting of the same size and orientation, keeping the hands and irrigation hose unchanged. Once I snap the green clip onto the plastic connector, add a small brown rabbit sitting next to the person’s boots, its nose twitching toward the hands holding the connector. Only after I remove the tube handle should the air shift to a fine, warm dust haze rising from the dry soil, softening distant edges while keeping my hands crisp.

Before event: 0 frames (0.00s – 0.00s)  ·  Event: 40 frames (0.00s – 2.50s)  ·  After event: 41 frames (2.50s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: When I pick up the detached connector and examine it closely, replace the black plastic connector with a brass fitting of the same size and orientation, keeping the hands and irrigation hose unchanged. Once I snap the green clip onto the plastic connector, add a small brown rabbit sitting next to the person’s boots, its nose twitching toward the hands holding the connector. As soon as I tie the loose end of the drip line to a stake in the soil, shift the air to a fine, warm dust haze rising from the dry soil, softening distant edges while keeping my hands crisp.

Before event: 0 frames (0.00s – 0.00s)  ·  Event: 40 frames (0.00s – 2.50s)  ·  After event: 41 frames (2.50s – 5.06s)

fb4a1e90f60f25f14cb7d906f311e7a5_event0_p1

GT edit prompt: Add a small bronze compass resting on the edge of the green mat near the left hand, casting a natural shadow and matching the scene’s perspective.

Single Positive

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Following my moving the golf ball, add a small bronze compass resting on the edge of the green mat near my left hand, casting a natural shadow and matching the scene’s perspective.

Before event: 31 frames (0.00s – 1.94s)  ·  Event: 45 frames (1.94s – 4.75s)  ·  After event: 5 frames (4.75s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: After I adjust the position of the golf club in my hands, add a small bronze compass resting on the edge of the green mat near the left hand, casting a natural shadow and matching the scene’s perspective.

Before event: 31 frames (0.00s – 1.94s)  ·  Event: 45 frames (1.94s – 4.75s)  ·  After event: 5 frames (4.75s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: As soon as I pick up a tee from the ground and place it in the basket, change the color of the green mat to a deep navy blue while maintaining the same texture and shadow patterns. After I wipe my hands on my pants, add a faint glow of orange light around the golf ball as the person picks it up, simulating heat radiating from the ball. When I step back to reposition myself further from the mat, replace the weave pattern on the green mat with a smooth, rubberized texture while preserving its size and depth. Following my moving the golf ball, add a small bronze compass resting on the edge of the green mat near my left hand, casting a natural shadow and matching the scene’s perspective.

Before event: 31 frames (0.00s – 1.94s)  ·  Event: 45 frames (1.94s – 4.75s)  ·  After event: 5 frames (4.75s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: As soon as I pick up a tee from the ground and place it in the basket, change the color of the green mat to a deep navy blue while maintaining the same texture and shadow patterns. After I wipe my hands on my pants, add a faint glow of orange light around the golf ball as the person picks it up, simulating heat radiating from the ball. When I step back to reposition myself further from the mat, replace the weave pattern on the green mat with a smooth, rubberized texture while preserving its size and depth. After I adjust the position of the golf club in my hands, add a small bronze compass resting on the edge of the green mat near the left hand, casting a natural shadow and matching the scene’s perspective.

Before event: 31 frames (0.00s – 1.94s)  ·  Event: 45 frames (1.94s – 4.75s)  ·  After event: 5 frames (4.75s – 5.06s)

fbed8b80c3247cab2dd70255e1580681_event2_p0

GT edit prompt: Replace the dough with a smooth, matte terracotta sculpture of a curled sea lion, fitting exactly in my gloved hands.

Single Positive

Source

GT Target

EgoEdit

Edit 1.5Visual 1.5BG 1.5Total 4.5

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 2.2Visual 2.2BG 2.2Total 6.8

EgoPlay-Causal

Edit 1.5Visual 1.5BG 1.5Total 4.5

Event prompt: Only after I pick up the dough should it be replaced with a smooth, matte terracotta sculpture of a curled sea lion, fitting exactly in my gloved hands.

Before event: 16 frames (0.00s – 1.00s)  ·  Event: 34 frames (1.00s – 3.12s)  ·  After event: 31 frames (3.12s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 0.0Visual 0.0BG 0.0Total 0.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 1.0Visual 1.0BG 1.0Total 3.0

Event prompt: After I adjust the digital scale display by pressing the buttons, replace the dough with a smooth, matte terracotta sculpture of a curled sea lion, fitting exactly in my gloved hands.

Before event: 16 frames (0.00s – 1.00s)  ·  Event: 34 frames (1.00s – 3.12s)  ·  After event: 31 frames (3.12s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 2.2Visual 2.2BG 2.2Total 6.8

EgoPlay-Causal

Edit 2.2Visual 2.2BG 2.2Total 6.8

Event prompt: After I scoop flour from the yellow container with my hand, change the person's одежда to a black short-sleeved uniform shirt with a small logo on the chest while keeping the blue gloves and workspace unchanged. Only after I pick up the dough should it be replaced with a smooth, matte terracotta sculpture of a curled sea lion, fitting exactly in my gloved hands. Once I slide a plastic bin toward the far edge of the table, change the color of the dough to a deep amber brown while maintaining the same texture and shadows. After I press the machine's start button with my thumb, change the background to a modern kitchen with light-colored cabinets and a tiled backsplash while keeping the dough and hands in the same position.

Before event: 16 frames (0.00s – 1.00s)  ·  Event: 34 frames (1.00s – 3.12s)  ·  After event: 31 frames (3.12s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 2.0Visual 2.0BG 2.0Total 6.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 2.0Visual 2.0BG 2.0Total 6.0

EgoPlay-Causal

Edit 1.0Visual 1.0BG 1.0Total 3.0

Event prompt: After I scoop flour from the yellow container with my hand, change the person's одежда to a black short-sleeved uniform shirt with a small logo on the chest while keeping the blue gloves and workspace unchanged. After I adjust the digital scale display by pressing the buttons, replace the dough with a smooth, matte terracotta sculpture of a curled sea lion, fitting exactly in my gloved hands. Once I slide a plastic bin toward the far edge of the table, change the color of the dough to a deep amber brown while maintaining the same texture and shadows. After I press the machine's start button with my thumb, change the background to a modern kitchen with light-colored cabinets and a tiled backsplash while keeping the dough and hands in the same position.

Before event: 16 frames (0.00s – 1.00s)  ·  Event: 34 frames (1.00s – 3.12s)  ·  After event: 31 frames (3.12s – 5.06s)

fc0ed5aaaeb6cc9d86c4b1f1b842fe0d_event0_p2

GT edit prompt: Replace the white rectangular object circled in green with a small vintage radio wrapped in brown leather, matching its size and orientation.

Single Positive

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 1.5Visual 1.5BG 1.5Total 4.5

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Upon taking a tin, replace the white rectangular object circled in green with a small vintage radio wrapped in brown leather, matching its size and orientation.

Before event: 27 frames (0.00s – 1.69s)  ·  Event: 33 frames (1.69s – 3.75s)  ·  After event: 21 frames (3.75s – 5.06s)

Single Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: After I pick up the green toolbox from the trunk floor, replace the white rectangular object circled in green with a small vintage radio wrapped in brown leather, matching its size and orientation.

Before event: 27 frames (0.00s – 1.69s)  ·  Event: 33 frames (1.69s – 3.75s)  ·  After event: 21 frames (3.75s – 5.06s)

Multiple Positive and Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: Upon taking a tin, replace the white rectangular object circled in green with a small vintage radio wrapped in brown leather, matching its size and orientation. As soon as I pull out the orange power drill and place it on the ground, remove the green hose from the trunk contents, ensuring the surrounding tools and materials are unaffected and the surface remains seamless.

Before event: 27 frames (0.00s – 1.69s)  ·  Event: 33 frames (1.69s – 3.75s)  ·  After event: 21 frames (3.75s – 5.06s)

Multiple Negative

Source

GT Target

EgoEdit

Edit 3.0Visual 3.0BG 3.0Total 9.0

VLM-guided

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay

Edit 3.0Visual 3.0BG 3.0Total 9.0

EgoPlay-Causal

Edit 3.0Visual 3.0BG 3.0Total 9.0

Event prompt: After I pick up the green toolbox from the trunk floor, replace the white rectangular object circled in green with a small vintage radio wrapped in brown leather, matching its size and orientation. As soon as I pull out the orange power drill and place it on the ground, remove the green hose from the trunk contents, ensuring the surrounding tools and materials are unaffected and the surface remains seamless.

Before event: 27 frames (0.00s – 1.69s)  ·  Event: 33 frames (1.69s – 3.75s)  ·  After event: 21 frames (3.75s – 5.06s)