How to Batch-Produce AI Ad Creatives: NadouPro Workflow for Ad Images and Videos

Ad materials are often consumed much faster than creative ideas can be produced. The same product needs to be placed across different channels, audiences, and campaign periods, and each version may need a new background, composition, or opening shot. Producing every version through traditional shooting puts heavy pressure on both cost and time.

This is where AI ad materials are valuable: a single product image can be expanded into a full set of image ads, or a text concept can be turned directly into a video ad with sound. NadouPro is iQIYI's professional filmmaking AI Agent platform, covering end-to-end generation capabilities from images and video to audio, making it well suited for this kind of batch ad production. Below, the complete workflow is explained through two tracks: image ads and video ads.

Before You Start, Separate What Can Change from What Cannot

Before generating anything, set a baseline: which details can be freely tested, and which must match the physical product. In general, backgrounds, lighting, camera position, and composition can be explored boldly, but the product's shape, color, and packaging text must stay consistent with the reference image. The clearer this baseline is, the easier it will be to review the results later.

Image Ads: Start from One Product Image

If you already have a product image, use Image Fusion to transform it. Upload the product image as a reference image, select Image Fusion mode in the Image node, and describe the parts you want to change—a different background, adjusted lighting, or a new placement angle—while clearly stating which details must remain unchanged, such as the bottle shape, main color palette, and text on the packaging.

If you do not have materials yet, use Text to Image to explore concepts. In the prompt, clearly state what the product is, what scene it is placed in, how the composition and lighting should work, and what overall visual style it should lean toward. This step is suitable for quickly producing concept drafts and seeing which visual direction is worth developing further.

Choose the model according to the material type:

• If the image needs a cinematic feel, or if it needs to arrange a large amount of text information, choose Seedream 5.0 Pro, which is strong at cinematic images and high-density graphic-text design;

• If the packaging contains fine text and requires an ultra-realistic texture, choose Qizhi-IGI 2, which accurately restores text and layout;

• If you need to edit multi-image combinations based on a single image, choose Seedream 4.5, which has strong multi-image editing capability and stable text generation;

• If you only want to quickly test composition directions, choose Seedream 5.0 Lite or Qizhi-IGI 2.5 Flare, which generate images quickly and offer good value.

Plan the output size in advance. When you need high-resolution large images, use the image ultra-HD capability to scale the image to the resolution required for placement. When composing the image, leave blank space for later titles and selling-point copy; do not let the product fill the entire frame. After generation, compare each image against the reference image to check color, proportions, and packaging details.

Roll out a series of materials in batches. When creating a full set of ads, build around the same product reference image: vary the background, composition, and shot size while keeping the product itself unchanged, so the whole set feels naturally unified. For fashion materials, you can use Image Fusion with a model image together with clothing and accessory reference images to test different outfit combinations. After generation, remember to compare against the real items to verify fabric details.

Video Ads: Make the Image Move and Sound

There are two entry points for video ads. Choose based on the materials you have:

• If you only have a text idea: use Text to Video mode and directly describe what happens in the frame;

• If you already have a product image you are satisfied with: use Image to Video or First/Last Frame mode to make the static image move.

Generate sound at the same time. The video models in NadouPro all support audio-video synchronization, generating sound along with the visuals. In the prompt, clearly write the visual action, how the camera moves, and what should be heard—ambient sound, background music, or character dialogue. If there are lines, simply mark the speaker and the line content, and the corresponding sound will be produced as the character speaks.

Use Storyboard for multi-shot ads. A complete ad usually contains more than one shot: an opening introduction, a product close-up, a usage demo, and a final end frame. You can use Storyboard to split the idea into shots with one click; the system automatically identifies characters, scenes, and props, then generates storyboard videos in batches, which can finally be assembled in Clip. You can also directly choose video models that support multiple shots, such as WAN 3.0 (smart multi-shot, ultra-long output), MiniMax H3 (multiple shots in one output), and Happy Horse 1.1 (smart multi-shot), to generate multiple shots at once.

Keep consistency across shots. The product and spokesperson should keep the same face and the same packaging across multiple shots, so use Multi-reference mode to lock the subject's appearance. When the voice-over needs to run through the full video, select a voice from the voice library in your assets and use Voice Dubbing to generate it consistently, avoiding different voices in each shot.

Taking a typical scenario as an example, the prompt for a perfume ad video could be written like this:

By the window of a Paris apartment, golden afternoon sunlight filters through the blinds and falls on a perfume bottle. The light spots move slowly, with gentle French piano music in the background. The camera slowly pushes in from scattered rose petals, shifting focus to the cut edges of the bottle. A female voice-over slowly begins, introducing the character of this perfume. The camera orbits slowly around the bottle, capturing the changes of light and shadow on the golden lettering. Finally, the camera pulls back and holds: the perfume bottle stands on a velvet base, with a faint view of a Paris street outside the window.

After generation, always watch the full video from beginning to end: the product may look correct in the first frame, but its proportions may quietly change once it starts moving; the lines also need to match the lip movements and actions. If something is wrong, revise only the description for that segment instead of overturning everything and starting over.

Material Ideas for Three Common Ad Placement Scenarios

Test openings for social ads. Use the same product reference image to create two versions of the opening shot: one Close-Up and one Full Shot, while keeping all other conditions the same so the opening becomes the only variable. Compare data at a small scale first, then decide which direction to scale.

Create demos for e-commerce ads. Use the product image to explore new display backgrounds, or generate a usage demo short video of a dozen or so seconds. For clothing materials, provide both the person image and the clothing reference image; after generation, verify clothing details and product proportions before creating a dynamic version.

Build series for seasonal and event marketing. Develop several holiday or seasonal scenes around the same product, and use reference images to unify the color tone and overall style, forming a visually coherent series of images. When video is needed, develop short scenes from text concepts, or choose a satisfactory image and turn it into video with Image to Video. Adapt the copy and composition separately for each placement position, and check every version before publishing.

Four Checks Before Launch

First, check the promotional claims. AI-generated demo visuals may imply functions the product does not have. Verify that the information conveyed by the visuals matches the actual product, make sure performance claims have a basis, and do not package fictional people or AI images as real users.

Second, check material authorization. Reference images, music, and on-screen likenesses all need legal authorization. Obtain permission before using a real person's portrait or voice, and keep authorization records.

Third, check labeling requirements. Understand the ad platform's labeling requirements for AI-generated content, and label it as required when necessary. Labeling cannot be used to endorse false advertising.

Fourth, check the final version. Verify prices, campaign dates, terms, and landing pages one by one; check packaging text, subtitles, and voice-over completely; preview every crop size once—the small text that looks clear in the original image may become unreadable at a smaller size.

Frequently Asked Questions

Q: Can AI make a complete ad?
It can produce image and video materials, but copy, selling points, and final review still need human oversight. AI is responsible for quickly visualizing ideas, while the team is responsible for setting the direction and maintaining quality.

Q: If I only have one product image, how do I make an AI ad?
For image ads, use Image Fusion: upload the product image and describe changes to the background, lighting, and composition. For video ads, use Image to Video or First/Last Frame, and describe the visual action and sound. In both formats, you must go back and check the product details.

Q: Can AI ad materials be used directly for real ad placement?
Yes, provided they can pass review. Having enough resolution is only the foundation; you also need to check product accuracy, text information, visual consistency, and the specific requirements of the placement platform.

Q: How can I keep the product consistent across different materials?
Always use the same product reference image for Image Fusion or Multi-reference generation, and clearly state in the prompt which details should remain unchanged (shape, color, packaging text). After generation, compare and inspect each image one by one.

Q: Are there any tips for batch-generating AI ad materials?
Set the baseline before scaling: fix the product subject and main color palette as constants, and batch-expand backgrounds, compositions, and shot sizes as variables. Whenever you get several satisfactory versions, save the reference images, then continue deriving later variants from the versions you like to improve efficiency.

Q: Are AI ads compliant?
There is no issue with using AI to make ads in itself, but you need to follow the rules of the placement region and platform: keep promotional claims truthful, ensure materials are authorized, and add AI labels as required.

Explore more features in NadouPro: NadouPro

Contains AI-generated information that may be incorrect. Please verify independently.