Wan 3.0 Document to Video is one of the most unusual additions to Alibaba’s latest video model. Instead of starting only from a text prompt, image, video, or audio reference, Wan 3.0 can use documents and webpages as creative source material. That means a presentation, report, spreadsheet, PDF, Markdown file, or public webpage can help define what the generated video should communicate.
Alibaba calls this broader capability “Everything to video” — a full multimodal reference system that, according to Alibaba Cloud’s official Wan 3.0 announcement, extends beyond text, image, audio, and video inputs to also read documents such as DOC, XLS, PPT, PDF, TXT, Markdown, Apple formats, and web pages. The goal is to turn lessons, tables, plans, links, and other information-heavy material into video rather than forcing users to manually rewrite every source into a long prompt first.
That makes Wan 3.0 Document to Video particularly interesting for product explainers, training materials, business reports, educational content, presentations, and data-driven videos.
XMK offers access to Wan 3.0’s current text, image, reference-media, 1080P, synchronized-audio, and long-form video workflows. Its publicly visible interface currently emphasizes text, image, video, and audio references, so document-input availability should be checked live, since Wan 3.0 remains in preview and its feature set can change.
What Is Wan 3.0 Document to Video?
Wan 3.0 Document to Video is a reference-driven workflow in which a document or webpage becomes part of the information used to generate a video.
A traditional text-to-video workflow might begin with:
Create a professional 30-second product explainer showing three benefits of a new analytics platform.
That tells the model what kind of video to make, but it does not contain the actual product information.
With Wan 3.0 Document to Video, the source material can instead come from something such as:
a PowerPoint sales deck,
a PDF product brief,
an Excel spreadsheet,
a Word report,
a Markdown document,
a Keynote presentation,
or a public webpage.
The prompt then becomes creative direction rather than the entire source of truth.
For example, instead of rewriting a 15-slide presentation manually, the user can provide the presentation and tell Wan 3.0:
Create a concise 30-second executive explainer from this presentation. Focus on the three most important findings, keep all numbers faithful to the source, use clear business narration, and end with the recommendation from the final section.
That distinction matters.
The document provides information.
The prompt provides direction.
This is the core idea behind Wan 3.0’s document-to-video mode.
What Documents Does Wan 3.0 Support?
Alibaba Cloud’s official announcement lists document types including DOC, XLS, PPT, PDF, TXT, Markdown, Apple formats (Keynote, Pages, Numbers), and webpages as supported reference inputs.
Input | Typical Formats | Good Use Cases |
|---|---|---|
Word document | DOC, DOCX | Reports, articles, product briefs, scripts |
Spreadsheet | XLS, XLSX | Metrics, financial summaries, survey data |
Presentation | PPT, PPTX | Sales decks, training slides, product launches |
White papers, reports, brochures, research | ||
Plain text | TXT | Scripts, notes, transcripts |
Markdown | MD | Documentation, structured articles, technical briefs |
Apple documents | Keynote, Pages, Numbers | Presentations, reports, spreadsheets |
Webpage | Public URL | Product pages, articles, landing pages |
Per Alibaba Cloud’s announcement, document input is capped at one file or link, up to 100MB and 50 pages per request — a single document or a single public webpage, not both combined. Public webpages need to be accessible without a login.
These limits make this mode better suited to a focused report, presentation, lesson, product document, or webpage than to an entire company knowledge base.
Wan 3.0 PDF to Video
Turning a PDF into video is probably the most obvious document workflow.
PDFs often combine:
text,
tables,
charts,
images,
diagrams,
headings,
and structured sections.
That makes them useful source material for explainer videos.
A 25-page product white paper, for example, could become a short overview covering only the three ideas most relevant to a customer.
The important mistake to avoid is asking Wan 3.0 simply to:
Turn this PDF into a video.
That gives the model very little guidance about what matters.
A better PDF-to-video prompt defines:
the audience,
the sections to prioritize,
what factual information must remain exact,
the desired duration,
the visual style,
and the ending.
Wan 3.0 PDF to Video Prompt
Use the attached PDF as the factual source for a 30-second explainer aimed at first-time customers. Focus only on the problem described in the introduction, the three product benefits in the middle section, and the final recommendation. Do not invent statistics or claims. Visualize the most important metric with a clean animated graphic. Use professional narration, restrained motion graphics, clear information hierarchy, and a simple closing product shot.
This works better because Wan 3.0 knows both what to read and what to make.
Wan 3.0 PowerPoint to Video
PowerPoint may be an even more natural fit for document-driven generation.
A presentation already contains a rough narrative sequence.
Slide 1 introduces the topic.
Slides 2–4 explain the problem.
Slides 5–7 introduce a solution.
Later slides provide evidence.
The last slide usually contains a recommendation or CTA.
That structure is much closer to video storytelling than an unstructured document.
Wan 3.0 PowerPoint to Video Prompt
Use the attached presentation as the narrative source. Create a 30-second product-launch video for a business audience. Open with the problem from slides 1–2, introduce the product using slides 3–5, visualize the two strongest proof points from the data slides, and finish with the recommendation on the final slide. Preserve product names, numerical claims, and labels exactly. Use clean technology visuals, smooth transitions, confident narration, and synchronized interface sound effects.
This approach can also work for:
pitch decks,
onboarding presentations,
employee training,
courseware,
conference presentations,
product demonstrations,
and marketing campaigns.
Alibaba explicitly positions educational materials and business information as intended scenarios for its document-reference capability.
Wan 3.0 Spreadsheet to Video
Spreadsheets create a different challenge.
A spreadsheet usually does not contain a ready-made story.
It contains data.
So the task is not simply converting cells into moving graphics. The model needs to understand which data deserves attention, and the prompt has to supply that judgment.
Wan 3.0 Spreadsheet to Video Prompt
Use the attached spreadsheet as the only source for all numerical values. Create a 25-second quarterly performance summary. Identify the three largest year-over-year changes, show each as a simple animated chart, and explain whether the value increased or decreased. Preserve every percentage and category label exactly as written. Do not invent causes that are not contained in the source. Use concise professional narration and a clean dashboard-inspired visual style.
Notice the phrase:
Do not invent causes that are not contained in the source.
That matters because data extraction and interpretation are different tasks.
A spreadsheet may show revenue rose 18%.
It may not explain why.
A document-to-video workflow should not turn missing context into an invented explanation.
Wan 3.0 Webpage to Video
Wan 3.0 can also use a publicly accessible webpage as reference material — Alibaba includes web pages directly among its supported document-style inputs.
Possible webpage-to-video use cases include:
turning a product landing page into a product overview,
summarizing a blog post,
converting an event page into a promo,
creating a video overview of documentation,
or transforming an educational page into a narrated short.
Wan 3.0 Webpage to Video Prompt
Use the referenced webpage as the source of truth. Create a 20-second vertical product explainer for social media. Introduce the product’s primary purpose, show the three features emphasized on the page, and finish with its main use case. Do not add features or pricing that are not present on the source page. Use fast but readable pacing, clean UI-inspired graphics, synchronized narration, and a strong opening visual in the first three seconds.
The webpage normally needs to be publicly accessible without account authentication for the model to retrieve it successfully.
How to Turn a Document into Video with Wan 3.0
The best workflow starts before generation.
Step 1: Clean the Source Document
Do not automatically upload the longest possible document.
Remove:
obsolete pages,
duplicate slides,
irrelevant appendices,
internal comments,
unfinished data,
and information that should not appear in the video.
A focused 10-page source is often easier to direct than a 45-page document containing several unrelated topics.
Step 2: Decide on One Video Goal
Ask:
What should the viewer understand after 30 seconds?
That answer should guide the prompt.
Possible goals include:
explain three product benefits,
summarize quarterly performance,
teach one concept,
introduce a new service,
highlight survey findings,
or turn a presentation into a client-facing overview.
Trying to summarize an entire document usually produces weaker information hierarchy.
Step 3: Provide the Document or Public Page
Wan 3.0’s reference workflow can accept a single document file or a single public webpage as input, alongside its broader multimodal reference system, per Alibaba Cloud’s documentation. File and webpage inputs are not combined in the same request.
Step 4: Add Creative Direction
The document does not replace prompting.
Tell Wan 3.0:
who the video is for,
which sections matter,
desired tone,
visual approach,
duration,
narration style,
what information must remain exact,
and what information must not be invented.
A useful formula is:
Source + audience + key information + narrative order + visual treatment + audio + factual constraints + ending
Step 5: Choose Duration and Resolution
Wan 3.0 supports video output up to 30 seconds, with official pricing tiers of $0.05, $0.10, and $0.20 per generated second for 480P, 720P, and 1080P respectively, according to Alibaba Cloud.
For experimentation, lower-resolution short tests are usually more economical than immediately generating a full 30-second 1080P version.
Step 6: Verify Every Factual Element
This is the most important step in any document-to-video workflow.
Check:
names,
statistics,
dates,
percentages,
labels,
chart values,
quotations,
product claims,
and narration.
Alibaba positions Wan 3.0 as capable of handling charts, data, interfaces, and other information-heavy content, but its own release notes also flag that audio texture and on-screen text rendering accuracy are still improving — a reminder that document input does not guarantee every factual or visual detail in the output will be exact.
For business, finance, education, legal, medical, or public-facing factual material, human review remains essential.
A Reusable Prompt Formula for Document Input
A reusable prompt formula is:
Use [DOCUMENT] as the source of truth. Create a [DURATION] video for [AUDIENCE]. Focus on [KEY SECTIONS]. Present the information in this order: [SEQUENCE]. Use [VISUAL STYLE]. Include [AUDIO/NARRATION]. Preserve [FACTS THAT MUST REMAIN EXACT]. Do not invent [UNSUPPORTED INFORMATION]. End with [FINAL MESSAGE].
Example
Use the attached product brief as the source of truth. Create a 25-second video for potential customers. Focus on the problem, the three core features, and the primary benefit. Begin with the customer problem, transition into a visual demonstration of each feature, and finish with the main outcome. Use clean modern product visuals, warm professional narration, and subtle interface sound effects. Preserve all product names and numerical claims exactly. Do not introduce features that are not in the document.
This structure gives the model both informational and creative boundaries.
5 Practical Use Cases for Document-to-Video
1. Product Launch Video
Input: A product deck containing screenshots, features, positioning, and launch messaging.
Output: A 20–30 second launch explainer highlighting the strongest value proposition.
2. Training Document to Video
Input: A PDF or presentation explaining a company process.
Output: A narrated micro-learning video focusing on the most important steps.
3. Quarterly Report to Video
Input: A report plus spreadsheet data.
Output: A concise executive video summarizing important movements and metrics.
4. Educational Presentation to Video
Input: A lesson deck.
Output: A short visual explainer introducing one concept with narration and animated diagrams.
5. Product Webpage to Video
Input: A public landing page.
Output: A social-ready product overview using information already published on the page.
These scenarios demonstrate why document input is more than another text-to-video mode. It moves AI video closer to existing business and information workflows.
What This Workflow Does Not Solve Automatically
Document input can reduce manual preparation, but it does not eliminate editorial judgment.
It Does Not Know What Matters Unless You Direct It
A 30-page report may contain dozens of valid facts. A 30-second video cannot communicate all of them. The prompt still needs to prioritize.
It Does Not Guarantee Perfect Numbers
If exact statistics matter, review them manually. A beautiful animation containing the wrong percentage is still wrong.
It Does Not Replace Video Editing
Generated video may still require:
subtitles,
brand graphics,
music adjustment,
exact cuts,
logo placement,
color correction,
or several clips stitched together.
It Does Not Turn Every Document into Good Storytelling
Documents are created for reading. Videos are created for watching.
A spreadsheet may need a clear narrative angle. A report may need aggressive summarization. A presentation may contain too much text.
The model can shorten part of the production chain, but the source still benefits from being prepared for the intended audience.
How Much Does Wan 3.0 Cost?
Alibaba Cloud’s current Wan 3.0 pricing is based on generated video duration:
Resolution | Price per Second | Approx. 30-Second Cost |
|---|---|---|
480P | $0.05 | $1.50 |
720P | $0.10 | $3.00 |
1080P | $0.20 | $6.00 |
Source: Alibaba Cloud’s Wan 3.0 announcement.
Document input does not change the fundamental logic: the major generation cost is tied to the resulting video duration and resolution.
For iterative document workflows, it therefore makes sense to test the information structure first before spending more on a final high-resolution render.
Third-party platforms can use their own credit or subscription pricing, so pricing outside Alibaba Cloud may differ.
Tips for Better Results
Keep the Source Focused
Remove pages that are unrelated to the final message.
Tell the Model What Not to Summarize
If a 30-page report contains only four relevant pages, say so explicitly.
Lock Important Facts
Use instructions such as:
Preserve all percentages exactly. Do not change product names. Use only claims contained in the source. Do not invent dates or statistics.
Write for Time, Not Page Count
A 30-second video has limited narrative capacity. Choose three important points rather than attempting to recreate every slide.
Separate Facts from Style
The document should determine what is true. The prompt should determine how it is presented. Keeping those roles separate produces clearer instructions.
Review the Final Video Against the Original
Keep the source open beside the result and verify the final narration, text, numbers, and visual claims before publishing.
Wan 3.0 Document to Video FAQ
Can Wan 3.0 turn a PDF into video?
Yes. Alibaba officially lists PDF documents as a supported reference input for Wan 3.0. The document can provide factual and structural context for a generated video.
Can Wan 3.0 turn PowerPoint into video?
Yes. PPT files are explicitly included in Alibaba’s Wan 3.0 document-input examples. Presentation decks are particularly suitable because they already contain a sequential narrative structure.
Can Wan 3.0 turn Excel spreadsheets into video?
Wan 3.0 supports spreadsheet-style document input, with XLS explicitly listed in Alibaba Cloud’s announcement.
Can Wan 3.0 turn a webpage into video?
Yes. Public webpages are included among Wan 3.0’s supported reference inputs. The page needs to be publicly accessible rather than protected by login authentication.
What is the Wan 3.0 document file limit?
Per Alibaba Cloud’s announcement, document input is limited to one file or link, up to approximately 100MB and 50 pages. Limits can change while Wan 3.0 remains in preview, so production workflows should verify the latest API or platform interface before relying on them.
Can Wan 3.0 preserve spreadsheet numbers exactly?
Wan 3.0 is designed to work with charts, data, and information-heavy sources, but generated factual content should still be verified against the original document. Do not assume that a generative video model guarantees perfect numerical transcription.
How long can a Wan 3.0 document video be?
Wan 3.0 supports video generation up to 30 seconds in a single output, with duration recommendation and extension features for longer narratives.
Does Wan 3.0 Document to Video generate audio?
Wan 3.0 can produce audio alongside video in the same generation workflow, but Alibaba’s own release notes note that audio texture is still being refined. Treat narration and sound output as a developing capability and review it before publishing, rather than assuming studio-quality audio by default.
Final Thoughts
Wan 3.0 Document to Video changes the starting point of AI video creation.
Instead of asking creators to compress an entire presentation, spreadsheet, PDF, report, or webpage into a single text prompt, Wan 3.0 allows that material to become part of the creative reference itself.
The biggest opportunity is not automatically making films from PDFs. It is shortening the distance between information that already exists and a video that communicates it.
That makes this workflow particularly useful for product teams, marketers, educators, sales teams, analysts, and creators working with information-heavy material.
The workflow still requires judgment. Source documents should be cleaned, the prompt should define priorities, and factual information must be checked against the original before publication.
But when those steps are handled carefully, document input can make Wan 3.0 much more than a conventional text-to-video generator.
Creators interested in the broader model can find Wan 3.0’s current long-form generation, native 1080P, synchronized-audio, text, image, and multimedia-reference workflows on XMK. As Wan 3.0 continues through preview rollout, check the live interface and Alibaba Cloud’s official documentation for the latest available input modes before starting a document-based project.