Start with the work. Trace it back to the maker and method.

A woman in a white silk robe walks half-way into thousands of golden light lines falling from the void, half her body in a plain underskirt and half covered by a gown woven from light. Shows how to write a dress as two materials coexisting on one body. Image 1 of Zhao Yifan's Dreamy Fashion set.

A champagne satin bodice turns at the waist into thousands of white and pale gold butterflies, looser further down until they become a free-flying cloud, with one butterfly resting on a fingertip. Uses a density gradient to write fabric as a living swarm. Image 7 of Zhao Yifan's Dreamy Fashion set.

A prompt that mimics the real, casual look of travel photos taken on a smartphone at a scenic spot.

Her pale gold chiffon dress is itself made of cloud, the hem and cloud surface losing their boundary as she spins, with the sun hidden below the clouds lighting upward and burning a gold edge. Shows how to make a hem and the scene's material one thing. Image 3 of Zhao Yifan's Dreamy Fashion set.

The frame edges fade to pure white while the center keeps saturated color, and her dress dissolves from the hem into a plain white cotton nightgown as she looks into the camera and says to remember this moment. Image 4 of Zhao Yifan's set; --hd is an old parameter.

A refined prompt for transforming uploaded photos into premium vertical posters, featuring realistic photography in the upper half and a stylized 3D paper-craft miniature scene in the lower half.

She walks barefoot across a mirror-like lake where the aurora reflection blurs sky and water, each step opening lotus-shaped concentric ripples, her silver-gray silk skirt melting into the water below the knees. Image 6 of Zhao Yifan's set.

She floats in pale pink water with eyes open, breathing normally, her white silk dress weightless as a cloud and her exhaled bubbles tiny glowing stars. Caustic light ripples over skin and dress like a water tattoo, binding person and fabric into one lighting system. Image 5 of Zhao Yifan's set; --hd is old.

A technical, realistic prompt for a future Tesla concept car, focused on believable engineering design and an automotive photography style.

A 90-degree profile close-up where wind sweeps across her cheek leaving watercolor traces of lavender, rose pink and honey gold that vanish after three seconds, with a petal fragment on her lip. Writes the wind as a brush that paints and erases. Image 2 of Zhao Yifan's Dreamy Fashion set.

Creates a photorealistic cinematic diorama of a sci-fi world physically emerging from an open book, with miniature figures and 3D typography.

A detailed prompt for generating a realistic TV broadcast screenshot of a high school baseball final, including specific scoreboard and channel-style text overlays.

A specific prompt designed for Nano Banana Pro to refine and smooth grainy or unnatural textures in AI-generated food images, particularly shaved ice, while preserving the original color scheme.

A cinematic prompt for a video of a child carefully sneaking out for a late-night adventure while their mother sleeps.

A photorealistic portrait of an East Asian woman in a traditional Japanese sailor school uniform sitting in a sunlit classroom.

A simple, creative prompt for first-person-view images of rare or unusual subjects, well suited to storytelling or concept art.

A comprehensive prompt for generating a seamless 10-second cinematic animation where a character catches a high-speed coffee mug and throws it back, focusing on realistic physics and character consistency.

A seamless-loop website hero animation demo generated with MiniMax H3. The character’s gunplay is mapped to eight web interactions—loading, hovering, clicking, scrolling, dragging, carousel navigation, submitting, and refreshing—each presented with corresponding ice-crystal UI feedback. The character and weapon designs are strictly locked to the reference image.

A Seedance 2.5 one-take video of a cliff wingsuit jump: the character leaps, opens the chute, skims the sea surface at speed dodging several yachts, then lands, removes the suit and ends by biting into a hot dog at a stall. The reference image locks body and face, and continuous follow-shots link the jump and landing.

A 15-second seamless-loop VFX video from MiniMax H3 with a reference-image character locked in place holding a long blade, using bullet-time orbits, kaleidoscope split screens and frame-grid montage so the blade shatters and then restores in reverse. It embeds animated Chinese and English titles.

A Seedance 2.5 game character-select interface animation that keeps the card carousel and skill-tree layout fixed while five spider anti-hero characters switch stances in turn. Each character's design is locked to its reference image and comes with its own skill effects.