Even with just a single product image on a white background, you can generate a complete demo video that showcases the product’s appearance, functionality, cleaning performance, and charging process.

In this tutorial, we’ll use a floor‑cleaning robot in white and mint green as an example. We’ll create a six‑panel product storyboard using PixPix, then write video prompts with an AI agent, and finally use the MiniMax H3 model to produce the final video.

Step 1: Upload the product image on a white background

Enter PixPix’s infinite canvas and upload the product image—on a white background—that you want to turn into a demo video.

To improve the stability of subsequent generation results, we recommend:

• A product image with a clean background

• The main body of the product should be fully visible and unobstructed

• A high‑resolution front view or side‑front perspective image

• An image where the product’s color, buttons, and structural details are clearly visible

The clearer the original product image, the higher the fidelity of the product in the resulting storyboard.

Step 2: Generate a six‑panel product demo storyboard

After uploading the product image, enter the instructions for creating a six‑panel demo storyboard for your video.

Figure 1 is the reference image of the product. Please generate a set of six‑panel product demo scenes set in a home environment, with a clean and tidy overall setting that conveys a natural, realistic lifestyle. The six panels should flow smoothly, follow clear logic, and highlight the product’s exterior details, core functions, and specific ways to operate its buttons. You may include close‑up shots and ultra‑close‑ups as needed to clearly show the buttons, operational processes, and functional effects, enabling viewers to intuitively understand the product’s features and how to use it. Each panel must adhere to a 9:16 vertical aspect ratio. The product must strictly replicate the shape, proportions, and details shown in the reference image, maintaining consistency across all frames; no elements in any frame should appear distorted.

After entering the instructions, select the image mode and wait for the AI to generate a set of six‑panel product demo scenes.

Step 3: Design the content of the six‑panel storyboard

In this case, each of the six panels serves a distinct purpose:

First, create a six‑panel grid to predefine the shot sequence for the entire video, thus avoiding narrative confusion, missing features, or inconsistencies in product presentation when generating the video directly.

Step 4: Use the Agent to generate video prompts

Once the six‑panel storyboard is complete, switch to Agent mode and have the AI generate corresponding video prompts based on each scene’s content.

Key elements of the video prompts should include:

• Specific actions of the product within the frame

• How hands interact with the product

• Camera movement techniques

• Changes in liquids and debris within the scene

• Consistency in the product’s appearance before and after

• The transition relationship between each shot and the next

Video prompts

Produce a 10-second, 9:16 vertical‑format, 2K, photorealistic commercial‑style product demonstration video for a household cordless floor scrubber, with natural spatial audio enabled.

Reference diagram relationship: The reference diagram is a six‑panel storyboard numbered 01–06. These six panels represent six full‑screen shots appearing in chronological order—not the final video’s grid layout. The finished video must play each shot in full screen, without displaying storyboard borders, numbers, grids, or text.

Product positioning: Throughout the entire video, the focus remains on the same unbranded household cordless floor scrubber. The main body features a matte warm white finish, accented by light mint green details and light gray structural components; the top houses a rounded D‑shaped handle with a circular power button fixed inside; a slender warm white metal upright pole; a semi‑transparent light cyan clean water tank mounted on the upper front half of the unit, and a semi‑transparent smoky gray wastewater tank installed on the lower half; at the base sits a low, rounded rectangular brush head, with a transparent window at the front housing a light mint green soft‑flocked roller brush, its leading edge fitted with a narrow white LED light strip, and a side‑mounted feature enabling edge cleaning. A warm white charging and self‑cleaning dock is included, with a light mint green self‑cleaning button attached to the side of the unit.

In all shots, the structure, proportions, colors, and placement of the handle, upright pole, clean water tank, wastewater tank, brush head, roller brush, LED lights, buttons, and charging dock must remain identical. The device must not transform into a vacuum cleaner, mop, or robot vacuum.

Scene setting: All scenes take place within the same bright, fresh, modern open‑plan kitchen–living room space. Creamy white walls and white cabinets, light oak flooring, beige carpet edging, mint green cushions, small potted plants, a clear glass vase, and sheer white curtains. Soft morning sunlight enters from the left side of the frame, with consistent lighting direction, furniture arrangement, and overall color palette. The visuals are crisp, clean, and relaxed—evoking the high‑end commercial aesthetics commonly seen on Taobao or JD.com product detail pages for cleaning appliances.

0.0–1.4 seconds | The only full‑view exterior shot

A full‑screen, 9:16 product hero shot. The complete floor scrubber stands upright on a warm white charging dock, fully visible from the D‑shaped handle to the rounded rectangular brush head, positioned at the junction of the kitchen and living room. The product occupies approximately 70% of the frame; the clean water tank has not yet been installed, and the power indicator is off. The camera slowly pans forward, as gentle morning light naturally glides along the warm white body, light cyan water tank, and transparent brush head.

This is the only instance throughout the entire video where the product’s exterior is shown in full. Subsequent shots may only highlight installation, operation, cleaning, and self‑cleaning details, but must never repeat this full exterior hero shot again.

1.4–2.8 seconds | Installing the clean water tank

Using the outline of the light cyan water tank as a visual cue, the scene transitions to a close‑up of the upper half of the floor scrubber. A pair of clean, natural adult female hands holds the translucent light cyan clean water tank filled with clear water, gently sliding it into its designated slot on the machine. The water level remains even, with no violent shaking or leakage.

Once the tank is fully seated in its slot, a subtle “click” sound signals that the latch has engaged properly. The fingers release naturally, leaving the tank securely in place. The frame must not once again display the entire floor scrubber in full view.

2.8–4.0 seconds | Pressing the power button

The shot shifts to a macro view of the power button inside the D‑shaped handle. A natural female index finger presses the round power button precisely; the button produces a realistic press-and-release action, with the edges glowing softly in mint green after activation, accompanied by a crisp, gentle startup tone.

Following the button press, the white LED light at the front edge of the brush head illuminates simultaneously, and the motor begins to run quietly and smoothly. No premature illumination, motor noise, or roller rotation should occur before startup.

4.0–6.1 seconds | Cleaning liquids and debris

As the startup tone plays, the camera switches to a low‑angle, ground‑level close‑up. The scene focuses solely on the lower half of the floor scrubber, the brush head, and the user’s legs clad in light beige house pants and clean slippers. She holds the handle naturally with both hands, steadily pushing the same floor scrubber forward.

In front of the light oak flooring lies a mixed wet‑dry stain composed of a small amount of light brown coffee liquid, clear water, and tiny biscuit crumbs. The light mint green roller brush begins to spin realistically; as the brush head contacts the stain, liquids and debris are continuously drawn beneath the transparent window. Areas untouched by the brush still retain their original stains; only after the brush passes does the floor become clean, even, and nearly dry.

The camera closely follows the brush head as it moves forward, ensuring that stains do not disappear prematurely, float away, or evaporate instantly. After processing, the floor must not be left with excessive water marks.

6.1–8.0 seconds | Roller brushing and edge cleaning

The camera naturally transitions to an extremely close‑up shot of the right side of the same brush head. Through the transparent window, one can clearly see the light mint green soft‑flocked roller spinning rapidly and evenly; a thin white LED light illuminates the wall‑adjacent floor area. The brush head moves forward, staying just a few millimeters away from the white wall baseboard, maintaining a safe distance.

Fine dust, biscuit crumbs, and minor water marks along the wall are completely sucked up as the brush head passes, creating a clear yet natural before‑and‑after comparison. The roller brush, transparent window, and brush head shape must remain exactly the same as in the previous shot; no product substitutions or additional second rollers are permitted.

8.0–10.0 seconds | Returning to the base for self‑cleaning

Using the brush head’s shape as a visual cue, the scene transitions to a close‑up of the lower half of the floor scrubber, now neatly placed into the same warm white charging and self‑cleaning base. The base and brush head align perfectly, with the machine standing stable and upright. First, a faint “click” sound indicates that the unit has successfully settled into its docking position.

A natural, feminine index finger presses the pale mint-green self-cleaning button on the side of the device. The button glows with a soft mint-green light, and the self-cleaning cycle can only start when the button is pressed. Inside the transparent brush-head window, the pale mint-green roller brush begins to spin and rinse within its base; a small amount of clean water evenly wets the roller, while dirty wastewater is drawn into the smoky-gray wastewater tank. The water flow must be controlled and realistic—no splashing outward, ensuring the surrounding floor remains perfectly clean.

The camera slowly pans in to reveal the transparent brush head, the rotating roller brush, the base interface, and the illuminated self-cleaning button, concluding with crisp, pristine product details.

Camera logic: a single, complete exterior shot → installation of the clean-water tank → pressing the power button to start → cleaning mixed wet-and-dry dirt → edge‑suction by the roller brush → returning to the base to initiate self‑cleaning. Each action must have a clear cause-and-effect relationship: the machine can only be turned on after the water tank is installed; the roller brush starts spinning only after the power button is pressed; stains disappear only after the brush head has passed over them; and the self‑cleaning function activates only once the machine returns to its base.

Scene transitions are crafted through clean, shape‑matching cuts that highlight the forms of the water tank, handle, brush head, and roller brush. Camera movements remain smooth and restrained, with continuous directional flow. Dissolving effects, flashing white, product melting, instant teleportation, flashy spins, or exaggerated sci-fi special effects are strictly prohibited.

Sound design: background music features lyric-free, fresh, modern domestic‑style tracks at approximately 110 BPM, incorporating gentle xylophone tones, crisp electronic beats, warm basslines, and subtle bright chimes. The opening includes faint ambient room sounds; natural plastic‑click noises accompany water-tank installation; a crisp “click” and brief startup tone signal power-on; during cleaning, steady low‑noise motor sounds, roller‑brush friction, gentle suction, and the faint sound of tiny debris being drawn in are heard; edge‑cleaning follows the brush head’s movement; a soft seating sound accompanies entry into the base; and self‑cleaning produces controlled roller‑brush rotation paired with circulating clean water. No dialogue or narration is used.

Strict prohibitions: the final video must not feature six‑panel grids, scene numbers, borders, subtitles, brand logos, or watermarks; it must not show a second model of the floor‑cleaning machine; repeated full‑view shots of the product are forbidden; the product must not appear distorted between shots; the positions of the clean‑water and wastewater tanks must not be swapped; no water leakage from the tanks; no changes to the brush head, roller brush, buttons, or base; the device must not transform into an ordinary vacuum, mop, or robot vacuum; the roller brush must not spin before the machine is powered on; stains must not vanish without the brush head passing over them; wastewater must not splash outward; large puddles must not remain on the floor; self‑cleaning must not activate before the machine returns to its base; extraneous fingers, deformed hands, or hands reaching through the water tank are prohibited; human leg proportions must remain consistent; backgrounds must stay static; camera shakes are forbidden; exposure jumps must be avoided; intense blue sci‑fi lighting is not allowed; cheap plastic textures and water flows defying physical logic are also unacceptable.

If you need to control each of the six scenes separately, you can also have the Agent generate an individual prompt for each frame and then produce the videos one by one.

Step 5: Select scenes and generate videos

Click on the desired scene in the six‑panel grid, then choose video mode and paste the Agent‑generated video prompts into the input box.

Next, complete the following settings:

1. Select the MiniMax H3 video model.

2. Set the video generation duration.

3. Set the output resolution.

4. Set the aspect ratio to 9:16.

5. Submit the generation task and wait for the results.

If you wish to create a full product demonstration video, you can generate each shot sequentially according to the six‑panel grid order and then finalize the assembly using video editing software.

How can you improve the stability of product demonstration videos?

Maintain consistent product appearance

The prompt should explicitly require the product to closely match the reference image, including:

• Product color

• Body proportions

• Water tank position

• Handle structure

• Button layout

• Base shape

If the product’s structure noticeably changes in a particular shot, it’s recommended to regenerate that shot separately rather than simply inserting it into the final video.

Each shot should showcase only one primary action

When a single shot includes multiple actions—such as removing the water tank, turning on the device, and cleaning the floor—it can easily lead to awkward hand movements or confusing action sequences.

A more stable approach is to break down these actions:

• One shot showing removal of the water tank

• One shot demonstrating pressing the power switch

• One shot illustrating stain removal

• One shot showing edge‑cleaning

• One shot depicting placement back onto the base

Clearly describe the before-and-after changes

In videos for cleaning products, ground‑level changes are the most crucial proof of performance. The prompt can explicitly state:

• Before cleaning: presence of liquid, debris, or stains

• After using the product: stains gradually absorbed

• After cleaning: the floor returns to a clean state

• The device moves naturally and continuously

This makes the cleaning effect much more evident.

Include close-up shots of specific details

Details such as buttons, the water tank, and the cleaning brush are best shown in close‑up or extreme close‑up. These detailed shots help viewers quickly grasp how the product operates and enhance the overall commercial appeal of the video.

Summary

The core workflow for creating product demonstration videos can be summarized as follows:

Upload a white‑background product image → Generate a six‑panel storyboard → Use an AI agent to draft video prompts → Select MiniMax H3 → Set duration, resolution, and aspect ratio → Generate each shot individually → Edit into a finished video

By planning shots in advance through the six‑panel storyboard, you can organize the product’s appearance, operating methods, and functional effects into a coherent presentation sequence. Even with just a single white‑background product photo, you can swiftly produce a vertical‑format product demo video suitable for e‑commerce promotion.