xPPTX turns Word documents, PDFs, images and mind maps into editable slide decks with speaker notes, and can also restyle decks you already have. Built by the team behind Docmee AiPPT.
Used this tool? Rate it
Used this tool? Rate it
xPPTX is a browser-based tool that converts documents you already have into editable slide decks. The site states its purpose in a single line at the top of the homepage: "AI Slides Generator - Make Presentations Instantly", followed by the promise that "In just one minute, Xpptx helps you turn Word documents into slides".
That framing matters, because it separates xPPTX from a large group of neighbouring products. Many AI presentation tools start from a blank prompt: you type a topic, the model invents an outline, and you get a deck about a subject the model guessed at. xPPTX is built around the opposite assumption — that the content already exists somewhere, in a report, a whitepaper, a PDF, a set of screenshots, or a mind map, and that the painful part is not inventing the argument but rebuilding it as slides.
The product exposes six conversion routes, each with its own page on the site: Word to slides, PDF to slides, image to slides, mind map to slides, outline to slides, and a beautification path for decks you already own. The last of these is the one worth noticing, because it means the tool is not only a generator. It also accepts an existing presentation as input and reworks it, which is a different job from creating one.
The product is delivered entirely through the web. There is no desktop application, no mobile app listed, and no plugin for PowerPoint or Google Slides described anywhere on the site. Third-party directory AlternativeTo lists the supported platform simply as "Online". Whatever you do with xPPTX, you do it in a browser tab, and the output is a file you then take elsewhere.
This is worth establishing early, because it changes how much weight the rest of the description deserves.
The footer of the xPPTX homepage carries a GitHub icon that links to github.com/docmee. That organisation is not a generic account. Its display name is "文多多 AiPPT" and its stated description is "文多多 AiPPT 开放平台" — the Docmee AiPPT open platform. It is based in China, points to docmee.cn/open-platform as its website, and hosts a set of integration SDKs: aippt-react, aippt-vue, aippt-api-python-demo, aippt-api-java-demo, and iframe embedding kits. In other words, xPPTX links from its own footer to an established AI presentation engine with a public developer platform.
Follow that thread and the picture resolves. Docmee, known in Chinese as 文多多, is operated by Shanghai MetaSign Intelligent Technology Co., Ltd. — 上海元符号智能科技有限公司. Its own about page publishes the details a shell site never publishes: an ICP filing number, 沪ICP备2022005812号-3, and a registered address at 上海市闵行区恒南路399号A栋 in the Minhang district of Shanghai. Both are verifiable in public Chinese registry records.
The link between that company and xPPTX is not inference. The xPPTX homepage carries an embedded About Us block naming "Shanghai MetaSign Technology Co., Ltd." and describing the company as "an AI technology company focusing on the field of intelligent office" that has "developed intelligent question answering robots, AI-generated slides and other products, and has served the financial, education, office and other industry fields". The founding team is described as holding doctorates from "Tsinghua University, Jiaotong University and other universities", with senior staff drawn from "Microsoft, Intel, Ctrip, Shunwang Technology (listed), Xiaoi Robot (listed)". That description matches, sentence for sentence, what Chinese technology coverage of Docmee reports about the same company.
So xPPTX is best understood not as an independent startup but as the international-facing brand of an established Chinese AI office-software company, running on the same lineage of presentation technology that powers Docmee domestically. The contact address on every page, support@metasigncloud.com, uses the MetaSign corporate domain rather than the product domain, which is consistent with that reading.
There is a discrepancy here that deserves stating plainly rather than smoothing over.
The About text embedded in the xPPTX site names a Shanghai company. The Terms of Use at /agreement/services name a different entity entirely: "Metasign Inc, a company incorporated in Wyoming, United States". The terms specify that they "are governed by the laws of the State of Wyoming, United States, excluding conflict-of-law principles", with disputes resolved by binding arbitration seated in Cheyenne, Wyoming. The Privacy Policy names the same Wyoming entity as data controller.
Neither document explains the relationship between the Shanghai company and the Wyoming company. The most ordinary explanation is a common and entirely legitimate structure — a Chinese company setting up a US entity to contract with international customers and to process payments through Stripe. But the site does not say that, and you should not assume it on the site's behalf.
What this means practically: the entity you are contracting with is the Wyoming company, because that is what the terms say, and terms govern. The entity whose engineering heritage and registry footprint you can actually verify is the Shanghai one. If your organisation has procurement rules about where a vendor is incorporated, where data is processed, or which legal system applies to a dispute, this is a question to raise with support before purchase rather than after. The answer is not published.
This section exists because it is one of the first things a careful buyer asks, and because the honest answer is partly negative.
xPPTX does not name the models it runs on. The homepage does not mention GPT, Claude, Gemini, DeepSeek, Qwen or any other model family — a direct search of the page source for those terms returns nothing. The Terms of Use acknowledge only that "The Services may rely on third-party providers, including payment processors, cloud and infrastructure providers, AI and content-processing providers", without naming any of them. The Privacy Policy is more specific about its subprocessors but stops short in exactly the same place: it names Stripe, Inc. for payments, Google Identity Services for sign-in, and Google Analytics, Google Ads and Microsoft Clarity for measurement — but no AI model vendor appears anywhere in the list.
What can be established comes from the parent company rather than the product page. Chinese technology coverage of Docmee reports that Shanghai MetaSign has conducted large language model research and training since its founding and has developed an industry-specific language model at the ten-billion-parameter scale. The company's own about page describes it as building "以行业大语言模型为基础" — on the foundation of an industry large language model. The xPPTX interface strings themselves reference "大模型配图", localised in the English build as "AI Model-Powered Image Matching", which indicates a model is doing image selection as well as text work.
Put together, this suggests a product built substantially on in-house models rather than a thin wrapper reselling somebody else's API. That is a meaningfully different risk profile from an anonymous reseller: the company has published registry details, a developer platform, open-source SDKs, and a research claim it attaches its own name to.
But the honest caveat stands. "Suggests" is not "states". The specific model behind any given generation is not disclosed, there is no published model version, and there is no commitment about whether the model might change. If your evaluation depends on knowing precisely which model processes your documents — and for some regulated content review processes it genuinely does — you will need to ask support directly, because the website will not tell you.
The six conversion routes are not variations on one feature. They handle materially different problems.
This is the flagship path and the one the homepage headline advertises. The relevant solutions page describes the behaviour precisely: "Upload a Word file and Xpptx quickly builds a structured slide deck that preserves chapter flow, key points, and core context."
The emphasis on preservation is the substance here. Converting a long document to slides is fundamentally a compression problem, and the failure mode of naive tools is that they either dump paragraphs onto slides verbatim or discard so much that the argument breaks. xPPTX claims to preserve headings, tables and key points while producing "editable slides pages with cleaner visual hierarchy and presentation-ready formatting".
The feature that does the most work in practice is speaker notes. The page states that "The AI document to Slides converter creates detailed speaker notes for each slide". If you are converting a twenty-page report into a twelve-slide deck, the material you cut has to go somewhere or you will not be able to speak to your own slides. Pushing it into the notes field is the correct answer, and it is the difference between a deck you can present and a deck that merely looks finished.
PDF is a harder input than Word and the site is candid about the ambition without being specific about the mechanism. The page claims the converter "can process large PDF files and dense content efficiently, including reports and reference documents with mixed layout patterns", and that "Xpptx identifies headings, paragraphs, and visual sections to rebuild slides with better structure". The stated philosophy is to "keep meaning, not just copy blocks".
What the page does not address is the case that matters most for anyone converting older material: scanned documents. A PDF containing a text layer is a structured document; a PDF containing photographs of pages is an image problem requiring optical character recognition. The site does not say whether scanned PDFs are handled, nor whether OCR runs at all on this path. If your source archive is scanned, test that specific file on the free tier before you commit money to it.
The image path takes screenshots, photographs of documents, or diagram captures and rebuilds them into structured slides. The homepage describes extracting text from complex layouts and reconstructing it. This is the route that most clearly implies optical character recognition even though the site never uses the term.
This route is unusually specific, and specificity is a good sign. The page names three accepted sources: "XMind, FreeMind, and Mubu links." Mubu — 幕布 — is a Chinese outlining and mind-mapping service, and its inclusion is another quiet indicator of the product's origins.
The conversion logic is described as mapping "branches to slide groups so your audience can follow logic, priorities, and dependencies", so the tree structure becomes the deck structure rather than being flattened. For anyone who plans in an outliner before writing, this is the shortest path from thinking to presenting.
The outline route accepts a structure you have written yourself and expands it into slides. It sits between prompt-based generation and document conversion: you control the skeleton, the tool supplies the flesh.
The beautification path is the one that most distinguishes xPPTX from pure generators, because the input is a finished presentation rather than a source document.
The page describes three categories of change. Text: "Text can be expanded for clarity or tightened for impact, while headings, key points, and emphasis are adjusted." Visuals: "Images, icons, and decorative elements can be removed, replaced, enhanced, or repositioned." Layout: the tool "improves spacing, alignment, color, and typography so the slides feel more consistent."
Critically, the result stays editable. The page states that "After content and visual cleanup, Xpptx keeps the deck editable so you can review, revise, and present with less rework", and that users can "Keep refining the result in PowerPoint, Google Slides, or your normal online editing workflow". A beautifier that returned a flattened, uneditable file would be close to useless for real work; this one does not.
Pricing is published openly at /pricing, without a login wall. That is worth crediting, because a significant number of comparable tools hide their numbers behind registration.
The free tier costs "$0" and includes "2 Slides generations" and "1 download". Two generations is enough to test whether the tool handles your document type, and not enough for anything else. Treat it as a trial, not a tier.
The Pro plan is "$15.99/month", or "$191.90 billed annually" with a stated 20% discount, and includes "14400 credits per year". Listed features are "Unlimited downloads", "No watermark export", "AI image matching", and "AI-powered smart layout".
The Ultra plan is "$27.91/month", or "$334.90 billed annually" at a stated 30% discount, and includes "28800 credits per year" — exactly double Pro's allowance. The only additional listed feature is "Faster support".
Separately, three one-time credit packs are sold, all described as never expiring: "$9.99" for "300 credits", "$29.90" for "1200 credits", and "$45.90" for "2400 credits". Non-expiring credits suit irregular usage — a few decks per quarter — better than a subscription does.
Payment runs through Stripe, described on the pricing page as "Secure payment powered by Stripe", accepting "credit cards, debit cards, and PayPal". Business buyers needing an invoice are directed to email support@metasigncloud.com with order and company details.
Here is the problem with all of the above, and it is not a small one.
The numbers tell you how many credits you get. Nothing on the public site tells you what a credit buys. There is no published rate for how many credits a Word conversion consumes, whether a fifty-page PDF costs more than a five-page one, whether AI image matching is billed separately from generation, or whether regenerating a deck charges again.
The official FAQ addresses this question and declines to answer it. Asked "How does the credit system work?", the response is: "Xpptx uses credits for paid usage. Depending on your order, credits may come from subscription credits, one-time credit packs, or promotional grants. The exact credit amount and included features are shown on the checkout and pricing cards at the time of purchase."
That explains where credits come from. It does not explain where they go. And because pricing is expressed in credits rather than in decks, the practical question — how many presentations does $15.99 actually produce — cannot be answered from published information at all. Pro's 14,400 credits per year could mean a hundred decks or a thousand.
The recommendation follows directly. Run the free tier first and watch what the balance does after a single conversion of a document representative of your real workload. That one observation converts the entire pricing table from abstract to concrete, and it costs nothing. Anyone quoting you a decks-per-month figure for this tool without having done that is guessing.
The refund position is consistent across both places it appears, which is more than can be said for many products in this category.
The pricing FAQ states: "Except where required by applicable law, paid fees are non-refundable. Please review the pricing details and plan scope at checkout before payment." The Terms of Use say the same thing: "Except where required by law, fees are non-refundable."
Cancellation is treated separately and equally plainly. From the terms: "Cancellation stops future renewal charges but does not retroactively cancel or refund charges already billed for the active billing period, except where required by law." Subscriptions "renew automatically at the end of each billing cycle until canceled."
One operational quirk is worth flagging because it will catch people out. Switching plans is not a single action. The FAQ states that "If you want to switch to another subscription plan, please cancel the currently active subscription first, then purchase the new plan", and the interface carries a matching message: "You already have an active subscription. Cancel it before subscribing to another plan." Upgrading from Pro to Ultra is therefore a two-step manual process, not a one-click upgrade, and there is no described proration.
Read together: this is a no-refunds product with auto-renewal on by default. That is a common arrangement and not in itself a red flag, but it does mean the free tier is the only risk-free evaluation you get, and the two-generation limit means you should plan that test carefully rather than burning it on a trivial file.
This section is longer than it would be for most products, because xPPTX publishes unusually few hard numbers about its own constraints. The absence is itself information a buyer needs.
Across every solutions page examined — Word, PDF, image, mind map, beautification — not one states a maximum file size, a maximum page count, a maximum slide count, or a maximum document length. The PDF page goes as far as claiming the converter handles "large PDF files and dense content", but "large" is not a number.
For short documents this is irrelevant. For the case where a document converter earns its money — the ninety-page annual report, the dense technical specification, the thesis chapter — it is precisely the information you need and cannot get. There is no way to know from the website whether your document will convert, truncate silently, or fail, and silent truncation is the outcome to fear because it is the one you might not notice until you are presenting.
The mitigation is the same as for credits: test your largest realistic file on the free tier and count the output slides against the source structure.
The PDF page describes its output in qualitative terms — "more precise", "more powerful", "more elegant" — and offers no accuracy benchmark of any kind. No figure is published for how reliably tables survive conversion, how often multi-column layouts are read in the correct order, or how equations, footnotes and captions are handled.
This is normal for the category and not a mark against xPPTX specifically. It does mean that fidelity claims should be treated as marketing until you have verified them against your own material, particularly if your documents contain the structures that break converters: nested tables, multi-column academic layouts, footnote-heavy legal text, and mathematical notation.
The interface strings promise "海量精美模板随意用", rendered in English as "Unlimited Access to Numerous Templates", and the free tier's restriction to one download implies templates are gated behind payment. But no template count appears anywhere on the public site, and there is no public gallery to browse before buying. You cannot assess whether the visual range fits your brand until you are inside the product.
Export is described unevenly. The Word path clearly produces editable PowerPoint output, and the membership feature list includes "Download Editable slides Source Files" alongside "No Watermark" export. But no page enumerates the full set of export formats. Whether PDF export exists, whether image export exists, and whether Google Slides export is offered are all unstated.
The watermark position is clearer and matters for the free tier: watermark-free export is listed as a paid feature, which implies free-tier downloads carry a watermark. The free tier's single download should be assumed to be watermarked unless you observe otherwise.
Restating this because it is the single most likely source of disappointment: no page says whether scanned or image-only PDFs are supported, and OCR is never mentioned by name anywhere on the site despite the image-to-slides feature strongly implying it exists. If your archive is scanned, this is the first thing to test.
xPPTX has essentially no third-party evaluation. Searches across Product Hunt, G2, Capterra and Trustpilot return no listing for the product — the results are all near-name collisions like XpertDox, XP-PEN, Projectum xPM and XEPOS, none of them related. Reddit and Hacker News contain no substantive discussion of the tool.
The one independent directory entry that could be retrieved, on AlternativeTo, records a single like, zero comments, zero reviews and zero listed alternatives. That same entry also contains an internal contradiction — it labels the licence "Proprietary (GPL-3.0)", which is self-cancelling and conflicts with the proprietary terms published on the official site. Two other directory listings could not be retrieved at all: both returned HTTP 403 through direct fetching, and one was additionally refused by local access controls, so nothing from either was used.
The consequence is straightforward. There is no body of independent user experience to consult, no aggregate rating, and no accumulation of complaints or praise. Every substantive claim in this article traces to the vendor's own pages or to verifiable corporate registry information, because nothing else exists. That is a normal position for a recently launched product — the site's structured data gives a publication date of 2025-10-17 — but it means your own testing carries all the evaluative weight that reviews would normally share.
A practical annoyance: "pptx" is a file extension, so searching for this product returns an ocean of unrelated PowerPoint utilities and format converters. If you are looking for support material, community discussion, or comparisons, expect the search results to be heavily polluted. Anchor on the domain xpptx.com rather than the name.
The terms are clear and generous on this point: "Subject to your compliance with these Terms, Metasign assigns to you any rights Metasign may hold in Output generated for your account." Users retain rights in their input and receive assigned rights in the output. Commercial use is permitted.
There is one condition attached that matters more than it first appears. Commercial use must comply with the licences of any third-party material incorporated into the output. Since the tool performs "AI image matching" and inserts visuals automatically, the images placed into your deck may carry their own licensing terms — and the site does not publish where those images come from or under what licence. For internal presentations this is a non-issue. For a deck going into a paid client deliverable, an advertisement, or a published report, verify the provenance of automatically inserted imagery before publication. That risk is created by the image-matching feature, not by the ownership clause, and the ownership clause does not neutralise it.
Notably, there is no self-contradiction here of the kind that appears in many tools of this class — the pricing pages do not promise commercial rights that the terms then withhold. The two documents agree.
The Privacy Policy describes retention in these terms: "Input, uploaded files, Output, and editing data is generally retained while needed to provide the requested workflow, preserve projects or account history, or until deletion through available controls or a verified request, subject to backups and legal obligations."
No number appears. There is no thirty-day window, no ninety-day window, no stated deletion schedule. Retention is bounded by purpose rather than by time, and deletion is available on request or through account controls. For most users this is unremarkable; for anyone uploading confidential or regulated documents, an open-ended retention period with no published maximum is a genuine compliance question rather than a formality.
The policy's position is conditional rather than absolute: "We do not publicly disclose or sell private Input or Output. Any materially different use of private Input or Output, including use to train a general-purpose AI model, will be governed by an additional product notice, setting, or agreement where required by applicable law."
Read carefully, this is not a flat commitment that your documents will never train a model. It is a commitment that such use would require a further notice, setting or agreement where the law demands one. The qualifier "where required by applicable law" does real work in that sentence. If you need an unconditional no-training guarantee — and many enterprise procurement processes do — this wording does not provide it, and you should seek written confirmation from support rather than relying on the published policy.
The Privacy Policy does not state which country or region hosts the servers. Given that the parent company is Shanghai-based, the contracting entity is registered in Wyoming, and the domain resolves through Chinese nameservers to an Alibaba Cloud address, the processing location is a reasonable question with no published answer. For teams subject to data residency requirements, this must be resolved with the vendor before any real document is uploaded.
To the site's credit, the third parties it does name are named plainly: Stripe, Inc. and its affiliates for payments, Google Identity Services for authentication, and Google Analytics, Google Ads and Microsoft Clarity for analytics and measurement. Microsoft Clarity in particular records session behaviour, which is worth knowing if you work with sensitive material on screen.
You regularly convert existing documents into decks. This is the core competence, and the speaker-notes generation is what makes converted decks presentable rather than merely produced.
You work from outliners or mind maps. Explicit XMind, FreeMind and Mubu support is more than most competitors offer.
You have inherited decks that need visual repair. The beautification path is genuinely differentiated, and the output stays editable.
You want to keep working in PowerPoint. The tool positions itself as a step in a workflow that continues in Microsoft or Google tools, not as a replacement editor to be locked into.
Your usage is irregular. The non-expiring credit packs fit sporadic use better than a subscription.
You need documented processing guarantees. No model disclosure, no data location, no retention ceiling, and no unconditional training opt-out. Regulated industries will not clear this on published documentation alone.
Your source material is scanned. Unaddressed on the site, and the most likely failure mode.
You need predictable per-deck costs. The credit consumption rate is not published, so budgeting is guesswork until you measure it yourself.
You want to buy on the strength of reviews. There are effectively none.
You need real-time collaboration or brand-locked template systems. Neither is described anywhere on the site.
Given how much is undisclosed, the free tier is not a formality — it is the only way to obtain the missing numbers. Two generations and one download is a tight budget, so spend it deliberately.
Pick the hardest document you realistically expect to convert, not an easy one. Convert it, then check three things: whether the output slide count is plausible against the source structure, whether tables and multi-column sections survived in the right reading order, and whether speaker notes were populated with the material that got compressed out.
Watch the credit balance before and after that single conversion. This is the single most valuable data point available and it resolves the entire pricing question.
Use the one download to confirm the export format and whether a watermark is applied.
If the source archive is scanned, test a scanned file specifically rather than assuming the PDF path covers it.
Only then decide between a subscription and a credit pack, and make that decision using the consumption rate you measured rather than the credit totals advertised.
xPPTX is a document-to-slides converter with a credible parent company, an honest core proposition, and a real gap between what it does and what it publishes about what it does.
The strengths are concrete. The conversion-first framing is the right one for people whose content already exists. Speaker-note generation solves the problem that makes most auto-generated decks unusable. Mind map support is specific and unusual. The beautification path addresses a genuine need that pure generators ignore. Pricing is published without a login wall, refund terms are consistent between the FAQ and the contract rather than contradicting each other, and output ownership is assigned to the user with commercial use permitted. The corporate lineage — a Shanghai company with a verifiable ICP filing, a public developer platform, open-source SDKs and a stated in-house model programme — puts this well clear of the anonymous wrapper sites that populate this category.
The weaknesses are equally concrete and all of one kind: things a buyer needs to know that are not written down. Credit consumption rates, file and page ceilings, template counts, export formats, scanned-document support, data residency, retention periods, and the identity of the underlying models are all absent. The relationship between the Shanghai company that built the technology and the Wyoming company that takes your money is unexplained. Independent verification is close to non-existent, and the one directory entry that could be read contains a self-contradictory licence field.
None of those gaps is evidence of bad faith. Collectively they mean the published record cannot support a purchase decision on its own. The free tier can — and unusually, a single careful test on it will answer more of the important questions than the entire website does. Run that test first, and treat everything advertised in credits as unquantified until you have watched the balance move.
A short note on positioning, since the category is crowded and the differences are easy to blur.
Against prompt-first generators such as Gamma or Tome, the distinction is the starting point. Those tools are strongest when you have a topic and no text. xPPTX is strongest when you have text and no time. If your raw material is a finished document, a prompt-first tool forces you to paste and re-explain content that already exists; the conversion-first framing avoids that step entirely.
Against Microsoft's own Copilot inside PowerPoint, the trade-off is access versus integration. Copilot lives where your deck already lives and inherits your organisation's tenancy and compliance posture; xPPTX is a separate web product with an unpublished data location and no stated model. But Copilot requires the corresponding Microsoft licensing, and xPPTX does not require you to be inside that estate at all.
Against straight format converters that turn a PDF into slides mechanically, the difference is that xPPTX claims to compress rather than transpose — condensing an argument and pushing the remainder into speaker notes rather than reproducing pages one-for-one.
The genuinely uncommon capability is the beautification path. Most competitors generate; comparatively few accept a finished deck and rework its text, imagery and layout while keeping it editable. If that is the job you actually have, the shortlist of alternatives is considerably shorter than the category's size suggests.
What xPPTX cannot compete on is documentation. Established competitors publish model details, data residency, retention windows and enterprise terms. That gap is not about output quality; it is about what a procurement team can verify before purchase.
It is a legitimate product with a verifiable corporate parent. The site's own footer links to the GitHub organisation "文多多 AiPPT" (Docmee), an established AI presentation platform with public SDKs, and the embedded About text names Shanghai MetaSign Technology. That company — 上海元符号智能科技有限公司 — publishes an ICP filing number, 沪ICP备2022005812号-3, and a registered Shanghai address, both verifiable in public records. xPPTX is best described as the international brand of an established Chinese AI office-software company rather than an independent startup or an anonymous reseller.
The site never says. No model family is named on the homepage, the Terms of Use refer only generically to "AI and content-processing providers", and the Privacy Policy names Stripe, Google and Microsoft Clarity as subprocessors but no AI vendor. The parent company is reported to have developed an industry-specific large language model at the ten-billion-parameter scale and describes itself as building on an in-house industry model, which suggests proprietary models rather than a resold API — but this is inference from the parent, not a statement about the product. Ask support if the answer matters to your evaluation.
The free tier is $0 and includes 2 slide generations and 1 download. Pro is $15.99 per month or $191.90 billed annually and includes 14,400 credits per year. Ultra is $27.91 per month or $334.90 billed annually and includes 28,800 credits per year, adding only faster support. One-time non-expiring credit packs are sold at $9.99 for 300 credits, $29.90 for 1,200, and $45.90 for 2,400. Payment is through Stripe.
This cannot be answered from published information. The site sells credits but never publishes how many credits a generation consumes, and the official FAQ explicitly declines to specify, saying only that credit amounts are shown at checkout. Because costs are denominated in credits rather than decks, the practical question of decks per month is unanswerable in advance. Measure your own consumption rate on the free tier before subscribing.
Word documents, PDF files, images, mind maps and outlines, plus existing presentations for restyling. Mind map support is the most specifically documented, naming XMind, FreeMind and Mubu links. What is not documented anywhere is whether scanned or image-only PDFs are handled, since OCR is never mentioned by name despite the image path implying it.
None are published. No solutions page states a maximum file size, page count, slide count or document length. The PDF page claims it handles "large PDF files and dense content" without quantifying "large". For long documents this is a real unknown, and silent truncation is the risk to check for by comparing output slide count against your source structure.
Yes, and you do. The terms state that Metasign "assigns to you any rights Metasign may hold in Output generated for your account", and commercial use is permitted. The one caveat is that output must comply with the licences of any third-party material it incorporates — and since the tool inserts images automatically through AI image matching without publishing their provenance, verify inserted imagery before using a deck in client-facing or published work.
Retention is open-ended. The Privacy Policy says files are "generally retained while needed to provide the requested workflow, preserve projects or account history, or until deletion" through account controls or a verified request, with no stated time limit. On training, the policy says private input and output are not publicly disclosed or sold, and that using them to train a general-purpose AI model "will be governed by an additional product notice, setting, or agreement where required by applicable law" — which is a conditional commitment, not an unconditional guarantee. Data storage location is not disclosed.
Generally no. Both the pricing FAQ and the Terms of Use state that fees are non-refundable except where required by applicable law, and cancellation stops future renewals without refunding the current period. Subscriptions auto-renew by default. Because of this, the free tier is your only risk-free evaluation, which is why it is worth spending its two generations on your hardest realistic document rather than a simple one.
Not in one click. The FAQ instructs users to "cancel the currently active subscription first, then purchase the new plan", and the interface repeats the requirement. No proration is described, so the timing of that switch is worth considering against your billing date.
