How it works

Preview mode and AI mode

The top bar always shows which mode you are in.

Preview mode (no AI). Slides are assembled by rules. Your instructions still decide the length, the level of detail, which sections get priority and which are left out (for example “skip the legal terms”). Section headings become slide titles and the document's own sentences become the bullets, each linked to its source. Tables and charts are copied from the document. Nothing is rewritten or translated, so slides can be wordier than an AI draft. Preview mode is free.

AI mode. A language model reads your instructions and the document, plans the presentation, and rewrites the material into presentation language: explanatory titles, concise points and speaker notes, in the language you ask for. It may summarise and rephrase, but every number must appear in the document; anything it suggests that the document does not state is marked as generated wording. Tables and chart data are always copied from the document, never written by the model.

Speaking your instructions

Select Dictate in the instructions box, allow the microphone when your browser asks, and speak. Select Stop and add to add the transcript to what you have typed, then review it. Nothing is sent until you select Create. If the microphone is blocked, what you typed is kept: allow the microphone in the browser's site settings and try again, or keep typing. The language you speak defaults to your browser's; change it with the link beside Dictate. To get the result in another language, say so in the instructions (“in Hindi”).

Dictation uses your browser's built-in speech recognition (Chrome, Edge and Safari). In Chrome and Edge the audio is sent to Google or Microsoft to be turned into text; we do not receive or store recordings. English, Hindi and several Indian languages are offered; whether a language is recognised, and how well, depends on the browser, and this has not been checked with a real microphone. Firefox has no built-in speech recognition. Where server transcription is switched on, the microphone there records your speech and sends it through our server to OpenAI to be turned into text; we do not store the recording. Where it is not switched on, the microphone is unavailable in Firefox: type instead.

One instruction, three formats

Write what you need and select Create: there is no outline to approve. Sliduno reads the format from your words: “presentation” or “slides” make a PowerPoint (the default), “report” or “PDF” a PDF report, “Excel” or “spreadsheet” an Excel workbook. The page says what it will make before you start, and every result can also be downloaded in the other two formats.

• PDF report: headings, findings, explanations as body text, tables that continue across pages, and charts drawn from their numbers. Text is selectable. Page references follow each point taken from the document.
• Excel workbook: a summary sheet; one sheet per table and chart, with numbers as numbers, empty cells where the document states nothing, formulas for counts, lowest and highest values, and editable Excel charts; the document's own tables in full; and a sources sheet.

After it is made, ask for changes in plain words in the editor: “make this shorter”, “compare the suppliers”, “add a price chart”, “explain this in Hindi”. Each change is a new version; earlier versions are kept. Asking only for another format (“give me the Excel version”) downloads the current version in that format without using AI.

Charts only use numbers found in your document: each value is checked against the table cell it came from, and counts are calculated by the software. A value the document does not give is left empty, never shown as zero. Text-heavy right-to-left or complex scripts in PDF reports may not be shaped correctly yet; the PowerPoint and Excel files are not affected.

What you get in PowerPoint

A standard .pptx file built on the chosen template's slide masters. Titles and points are real text you can retype, tables are native PowerPoint tables, charts are native charts with their data in an embedded workbook (Edit Data works), and speaker notes include the source pages or paragraphs used.

Supported chart types: clustered column, clustered bar, line and single-series pie. Other chart types found in a document are shown as tables of their data, never as a different chart. Charts that are only pictures stay pictures, and their numbers are not read. A picture from your document can be placed on its own slide, with a caption saying its values are not editable.

Export limits at a glance

• In: PDF and Word (.docx) only, up to 15 MB on this service. PDFs are read up to 40 pages; scanned pages are read with text recognition. PowerPoint files are accepted as templates, not as content to improve.
• Out: PowerPoint (.pptx), a PDF report, and an Excel workbook. The PDF is a report written from the same content, not a picture of the slides; a PDF copy of the deck itself is not available.
• Slides: title, list, card, figure-led, table, chart and picture layouts, chosen from what each slide holds. On the built-in designs a plain table also gets a chart of its own numbers. Your own template keeps its layouts. Comparison, timeline and process layouts are not available yet.
• Limits per account: every account has a limit on results in any 30 days, with or without AI, and on AI use in any 30 days. Editing and downloading what you already made stay free. AI stops when a plan ends.
• Charts: clustered column, clustered bar, line and single-series pie, as native editable charts. Other chart types become tables.
• Long text: text that does not fit is shrunk to a limit and then flagged in the editor, never cut off silently. Tables too tall for one slide continue on the next with the header repeated.
• Very long documents: text beyond about 240,000 characters is not used; you are told before generating.
• Maps and very detailed drawings: on such a page only the sentences are read, not its labels or tables. The result says which pages.
• Reports that take a long time to read: reading stops after about four minutes, keeps what it has, and names the pages it did not reach. A document needing more memory than is available is refused with a clear message, and nothing is charged.
• Tables in PDFs: figures are copied as printed. A table whose column headings cannot be told apart stays out of the slides and remains in the Excel download. Multi-level headings can read in part, so check them against the original.
• Length: preview decks keep to the slides you ask for. Tables stay whole, so a long table can take the place of other sections; the editor lists what was left out. AI decks can run a slide or two over when a table is split.
• Fonts: if a font is missing where the deck is opened, PowerPoint substitutes it and lines may wrap differently.

Tested so far

• The two built-in templates (16:9 and 4:3): files open in Microsoft PowerPoint for Windows (Microsoft 365) without a repair prompt; text, table and chart data were edited, saved and reopened.
• Exported files for every slide type on both templates pass Microsoft's Open XML SDK schema validation (Office 2016).
• PDFs with a text layer and Word .docx files, including Word charts with stored data.
• Web app in Chromium, Firefox and WebKit browser engines (desktop, plus a mobile viewport in Chromium), including automated accessibility checks.
• Scanned PDFs: English text is recognised with OCR on the hosted service, and slides using it say so. Check names, dates and numbers against the original.

Not yet verified

• PowerPoint for Mac, PowerPoint on the web, Keynote and Google Slides.
• Safari on macOS or iOS (WebKit tests are not the same as real Safari).
• Uploaded templates beyond basic structure checks: complex masters, unusual placeholders or embedded fonts may need manual adjustment.
• OCR for languages other than English, handwriting and low-quality scans.

Preview accuracy

The editor preview uses the same positions, font sizes and line breaks as the exported file. PowerPoint draws text and charts itself, so small differences in kerning, line height and chart styling remain. If a font is not installed where the deck is opened, PowerPoint substitutes another font and lines may wrap differently. Text that does not fit is shrunk to a limit and then flagged; it is never cut off silently.

Your documents

• Files are inspected before reading. Macro-enabled, password-protected and malformed files are refused with a reason.
• Links inside documents are kept as text. Remote templates, linked data, OLE objects and remote images are never fetched.
• Document text is treated as content, never as instructions to the system.
• Originals and downloads are kept in private storage for 30 days. Deleting a project or your account removes them from active storage immediately; backups expire on their own schedule.
• When AI generation is enabled, document text is sent to the AI provider (Anthropic) to write slides, subject to that provider's data retention terms. In preview mode, nothing is sent to an AI provider.
• Where it runs: the website on Cloudflare, document processing on Render, and the database and stored files on Neon.
• The operator's dashboard shows counts, costs and job failure messages, not the contents of documents or presentations. Whoever operates the service has technical access to the database and file storage.

Getting help

During early access, write to slidunosupport@gmail.com with the project name, the deck version and what you expected. Do not attach confidential documents to a support request. A first visit after a quiet spell can take about a minute while the service wakes.