In-Distribution Forcing for Long Video Generation at Test Time
(Supplementary Material)

You can freely control the playback speed and settings using the top bar!


Qualitative Comparison reported in the paper

We qualitatively compare 120-second videos generated by applying baselines onto two base models: Self-Forcing and LongLive. All methods in each row share the same seed for fair comparison.


Self-Forcing

Deep Forcing

Infinity-RoPE

MemRoPE

Ours

Prompt: A litter of playful golden retriever puppies popping their heads out of the snowy ground, their fluffy tails wagging and noses sniffling the cold air. They are covered in snow, with big brown eyes full of joy. The background shows a wintry scene with trees and a gentle snowfall. Wide-angle shot, medium shot of the puppies.



LongLive

Deep Forcing

Infinity-RoPE

MemRoPE

Ours

Prompt: A movie trailer featuring the adventures of a 30-year-old space man wearing a red wool knitted motorcycle helmet. He explores a vast salt desert under a blue sky, with vivid colors and cinematic style. Shot on 35mm film, the trailer captures him running across the desert, his helmet glinting in the sunlight. Dynamic camera movements and sweeping shots highlight his journey.




Additional Comparisons

We present further comparisons across 30s, 60s, 120s, and 240s durations using MovieGen prompts. As LongLive is tuned on long videos, we use Self-Forcing, trained only on 5s clips, as the base model so that every duration tests extrapolation beyond the training horizon. Our method consistently delivers high quality across all video lengths.

30s


Self-Forcing

Deep Forcing

Infinity-RoPE

MemRoPE

Ours

Prompt: Macro shot of a young woman's face filled with freckles, intently searching for something.


Prompt: Aerial view of Santorini during the blue hour, showcasing the stunning architecture of white Cycladic buildings with blue domes. The caldera views are breathtaking, and the lighting creates a beautiful, serene atmosphere with soft, warm hues. Camera slowly descends, capturing the intricate details of the buildings and the vast, tranquil sea below. Medium shot.


Prompt: Historical documentary style, bustling scenes of California during the gold rush. People in period clothing are panning for gold, mining, and trading in a dusty town square. Background shows old buildings, wagons, and horses. Authentic costumes and props, warm golden sunlight filtering through trees. Wide shot of the lively town and miners at work.


60s


Self-Forcing

Deep Forcing

Infinity-RoPE

MemRoPE

Ours

Prompt: Japanese animated film, a young woman stands on a ship, looking back at the camera with a thoughtful expression. She wears a traditional kimono with colorful patterns, her hair tied in a neat bun. The background shows waves and distant mountains, with a blurred horizon. Medium shot, slightly overhead view.


Prompt: Chimneys against the setting sun, casting long shadows, with warm orange and pink hues in the background. The scene has a nostalgic, golden hour feel. Long shot, capturing the entire skyline.


Prompt: A dynamic tsunami rushing through a narrow alley in Bulgaria, water surging and crashing violently. The alley walls loom tall and imposing in the background. Close-up, fast-paced shot.


120s


Self-Forcing

Deep Forcing

Infinity-RoPE

MemRoPE

Ours

Prompt: A cute rabbit wearing glasses sits on a chair, reading a newspaper. The rabbit has soft fur, big round eyes, and a small nose. It is sitting upright with its paws on its lap. The background shows a cozy living room with a fireplace and bookshelves. Soft lighting and warm colors. Close-up view, gentle reading motion.

Prompt: A giant humanoid made of fluffy blue cotton candy, stomping on the ground and roaring to the sky, with a clear blue sky behind them. Medium shot, dynamic view showing the creature's powerful stomps and fierce roar.


Prompt: Beautiful, snowy Tokyo city is bustling with activity. The camera moves through the lively city street, following joyful people enjoying the snowy weather and shopping at nearby stalls. Gorgeous sakura petals and snowflakes dance through the air together.


240s


Self-Forcing

Deep Forcing

Infinity-RoPE

MemRoPE

Ours

Prompt: Aerial shot of the ocean, a maelstrom forms in the water, swirling violently until it reveals the fiery depths below, with waves crashing and water vapor rising.


Prompt: A blooming garden with vibrant flowers in various colors and shapes. Soft sunlight filters through the leaves, casting dappled shadows. Close-up shots of different flowers close to the ground, with gentle wind swaying the petals. Pastel color palette, natural and serene.


Prompt: Animated 3D scene featuring a close-up of a short, fluffy monster kneeling beside a melting red candle. The monster has wide eyes and an open mouth, gazing at the flame with a sense of wonder and curiosity. Its innocent and playful expression conveys exploration and discovery. Warm colors and dramatic lighting enhance the cozy atmosphere. Medium shot, slightly overhead view.