- Customer: “Can I turn a product photo into a short video without changing the label?
- A Xelta prompt library workflow creates a well-reviewed answer to repeated questions.
- Take questions from support tickets, customer feedback and Search Console.
- Questions like ‘Why does the text on my packaging change when I animate it?
Introduction
Customer: “Can I turn a product photo into a short video without changing the label?” The help page describes how to upload an image and create a clip. However, it doesn’t explain what to do if the product name changes during the recording.
A Xelta prompt library workflow creates a consistent way to provide information like this. By linking each question to a known workflow, example and verification, the support and content teams can quickly refer back to previous answers.
A Xelta prompt library workflow creates a well-reviewed answer to repeated questions. Keep the original question, outline the action and restriction required to complete the task. Verify the workflow is up-to-date with the current documentation and create a clear answer with an example and verification. Repeat this process with other similar questions.

How to Break Down the User Question
Take questions from support tickets, customer feedback and Search Console. Keep the original question wording. Words like ‘with one image’, ‘without altering the label’ and ‘once exported’ help to understand what the answer should include.
Remove any personal customer information and any private project details. Save the example to the shared library.
Create a table for each question. Include what the user has to start with, what they want to achieve, key restriction and verification.
| Field | Example entry |
|---|---|
| Original question | Can I turn a product photo into a short video without changing the label? |
| Starting material | One approved product photo with a readable label |
| Desired output | A short product clip for a social post |
| Main constraint | The label must stay accurate and readable throughout the shot |
| Pass check | Compare the exported clip with the original photo, especially during camera movement |
This will help determine how detailed your answer needs to be. “Make sure the video looks good” will give the user little to do. Whereas ‘Stop the video at the point of greatest product movement and check the text is the same as the original image’ will give them clear guidance.
Group Similar Workflows Together
Questions like ‘Why does the text on my packaging change when I animate it?’ and “How do I keep the label readable in a product video?” can be grouped together. Both start with an approved image and both have a verification. Whereas “Can I make a product video without a product photo?” would be a different question.
Group questions by the action needed and decision to make. If the actions and verification differ too much then create a new entry.
Three Prompts to Add to Your Prompt Library
Understand the Workflow
Analyse the user question: [question]. Understand what they have, what they want to achieve, what they cannot ignore and what they need to decide. Suggest one workflow that can be done using only the confirmed capabilities: [notes]. Include what facts need verifying.
Verify: Check that your workflow matches the original question. If they have one image then ensure your answer will work with just one image. If they need the label to be accurate then ensure there is a verification for the exported footage. Highlight any facts that need verifying so the writer knows what to research.
Create a Standalone Answer
Answer the question [question] in 45-60 words. Use the [verified facts]. Outline what they have to start with, what action to take and the main restriction. Provide a clear workflow for the user, including one tangible verification and any fall back they can use. Highlight any facts that need verifying.
Verify: Ensure the first paragraph works on its own. Ensure the user knows what to try and what could go wrong. Place the specific verification next to the relevant action. For example, ‘Ensure the label isn’t distorted’.
Audit the Product Claims
Compare your draft to your [official documentation and verified notes]. List any claims made about the input, settings, features, plans, restrictions and quality of the output. Highlight if the claim is supported, unclear or not supported. Suggest more specific wording for any unclear claims or highlight as something to remove. Highlight any examples that could be seen as a test.
Verify: Check that the source supports the exact claim. If there is documentation to show they can upload images then the answer will include starting with an image. If there is a claim that the text won’t change throughout the whole video then this needs to be proved separately. Keep note of where the source was, what version of the prompt, date verified and who verified. If the team need to document model settings, reference assets and amendments then use the creative prompt guide.

Example: Product Video With Clear Label
This example is based on having an approved product image with clear front facing label text. The team want a short video with subtle movement. Start by using an AI video generator.
Published Answer
Start with your approved product image and request a slow zoom in. Compare the exported video to the original image. Ensure the label is clear whilst moving. If the text is shifting or not clear then create a slow zoom in within a video editor using the original static image.
Generation Prompt
Slow zoom in on provided product image. Face the package forward. Maintain shape, printed text, colours and logo. Use even lighting and ensure all of the label is visible.
First play the exported video at normal speed. Pause at the point of largest product in the frame or where there is most camera movement. Check that the product name, logo and any small text is clear. To reduce the movement you can start with reducing the movement. If you need the text to be precise then using a simple zoom in on the original static image will ensure the approved label is visible.
Different Inputs
Another customer may ask the same question with a partially obscured label. Here the issue is with the reference. Request a clear front facing image or use footage where the approved package can be seen. This will still fit within the same guide to show when the workflow can be used. It will also provide a specific answer for the support team to use if the uploaded image isn’t suitable.
Keep the Prompt Library Useful
Assign an owner to each entry. Keep the status simple such as draft, verified or needs reviewing. Keep the approved answer with the original question, source and last verified date. If there is a change to the upload options or tool labels then the owner can amend the relevant entries.
If there is a common example in production then AI creative workflows can be created to link the images, prompts and video. Keep the verification criteria with the prompt so new team members know what to check. Note why the amendment was made. “Reduced camera movement after the lower label text distorted”
What to Publish for AI Search
Publish the clear answer. Include the steps, example and fall back. Keep the different versions of the prompt and verification in the internal library. If there are multiple questions that have the Same Workflow then use these to improve the existing guide. If the user needs different inputs, steps or criteria then it makes sense to create a new page.
Google use standard Search criteria for AI Overviews and AI Mode. Ensure the page is crawlable, indexed and eligible to show a snippet. This means it is eligible to be shown as a supporting link. Whether it is shown depends on Google. Finally, as an editorial check ask if someone lands on the page they can complete the task without needing to refer to another guide.

Final Thought
Choose one question that has been answered multiple times. Keep the original question, verify the workflow and publish an answer with verification and fall back. Keep the verified answer so you can start with the evidence and reasons for each step when updating. For ideas on which example to use in production browse the AI video demos.
Frequently asked questions




