--- title: "Frequently asked questions" description: "Whether it invents citations, whether it works with seventeenth-century scans or video interviews, what happens to your data, whether it works offline, and other questions, answered plainly." url: https://scholaris.joseluissaorin.com/en/knowledge/faq markdown: https://scholaris.joseluissaorin.com/en/knowledge/faq.md lang: en alternate_es: https://scholaris.joseluissaorin.com/saber/preguntas.md updated: 2026-10-07 author: José Luis Saorín Ferrer (https://joseluissaorin.com) --- # Frequently asked questions > Whether it invents citations, whether it works with seventeenth-century scans or video interviews, what happens to your data, whether it works offline, and other questions, answered plainly. ## About citations ### Does it invent citations? No. No model writes a citation: author, year and page come from the document's record and from the anchor stored when it was read. When a model drafts an answer, it may only cite the passages it is given, and every note is checked against the text before you see it. If no passage supports something, the answer says so. In the 41-claim benchmark, invented citations were zero. See [Citations](https://scholaris.joseluissaorin.com/en/knowledge/citations.md). ### Is the page the PDF's or the book's? The book's: the number printed on paper, even when the PDF starts at the cover, the book uses roman numerals in the preface or skips plates. If the book prints no number, the physical position is cited in brackets, "p. [12]", instead of inventing one. See [Formats and anchors](https://scholaris.joseluissaorin.com/en/knowledge/formats.md). ### Which citation styles does it have? APA, Chicago (author-date and notes), MLA, Harvard, ISO 690 (author-date in Spanish and English, and numeric) and IEEE, in Spanish, English, French and Italian. It exports BibTeX, RIS and CSL-JSON, and inserts citations into a .docx with real Word footnotes. ## About what it reads ### Does it work with scanned seventeenth-century books? Yes, it is one of the cases it was built for. It reads pages with a vision model that keeps the original spelling ("dixo", "muger", u/v, ç) and leaves out running heads, catchwords and signatures; it recognises foliation by leaves ("23r", "23v"). So that those spellings show up when you search for "dijo" or "mujer", every passage carries a modernised-spelling layer used only for searching. On a 43-page play by Lope de Vega, the character error rate was 0.7 % and search recall went from 32 % to 100 %. See [Search](https://scholaris.joseluissaorin.com/en/knowledge/search.md#grafia). ### And with video interviews? Yes. It transcribes word by word, tells speakers apart and names them, cuts the text into 30-to-60-second stretches and cites the exact minute ("12:04-12:40"). Video frames are searchable too. A two-hour interview was ready in 70 seconds. YouTube is read from its address, without downloading the video. See [Audio and video](https://scholaris.joseluissaorin.com/en/knowledge/media-player.md). ### Does it read Latin? Yes. The search layer evens out u/v, i/j, æ/œ and the enclitics, and reduces words to their stem with the Schinke et al. stemmer, so "urbis" finds "Vrbs". ### Which formats does it take? PDF (digital and scanned), photos (JPG, PNG, HEIC…), EPUB, Word, ODT, RTF, HTML, Markdown, PowerPoint, Keynote, ODP, Excel, ODS, CSV, audio (MP3, WAV, M4A…), video (MP4, MOV, WebM…), web pages, YouTube, Vimeo and podcasts. The full list, with each one's anchor, is in [Formats and anchors](https://scholaris.joseluissaorin.com/en/knowledge/formats.md). ## About your data ### What happens to my data? It is stored on Cloudflare, in a space of its own for your account, and it is not used to train models. To read and search it, pages, audio and queries are sent to Google Gemini, Cloudflare Workers AI, OpenRouter and TypeSafe, each for a specific task; open bibliographic databases only receive titles, authors, ISBNs and DOIs. You can export everything and delete your account whenever you like. The detail, provider by provider, is in [Privacy](https://scholaris.joseluissaorin.com/en/knowledge/privacy.md). ### Can anyone else see my library? Only if you share it: by inviting someone by email or creating a public link, which you can protect with a password, set to expire or revoke. Those pages ask search engines not to index them. ### Can I use it offline? Yes, with the home version in offline mode: it reads pages with an open vision model (Qwen3-VL), embeds with EmbeddingGemma 2, transcribes with Whisper and judges citations with models running on your computer or your network, with no cloud key and nothing going out to the internet. It is slower (about 26 s per scanned page on an M4 Max Mac, against seconds for the whole book in the cloud) and somewhat less sharp, and without a network the bibliographic catalogues are not consulted. On top of that, opening and searching an .spdf that has already been read works with none of this, through the Python SDK. See [Self-hosting](https://scholaris.joseluissaorin.com/en/knowledge/self-hosting.md). ### Is it open source? The Python SDK is (EUPL-1.2). The rest of the code will be published under the same licence when its review is finished; there is no date yet. ## About using it ### Does it write my paper? No, and it will not. It searches, points, cites and verifies; thinking and writing remain yours. The "cite" function gives you back your own text with the citations in place, not a new text. ### Does it search the internet for papers? No: it works on what you give it. To discover literature you do not have yet, use an academic search engine and bring what you find here. The citation graph does tell you which works your books cite that you do not have. ### How much does it cost? There is a free plan (25 documents, 1,500 pages or minutes a month, 100 searches a day) and a Pro plan, whose price is shown in the app. The home version has no quotas. See [Plans](https://scholaris.joseluissaorin.com/en/knowledge/plans.md). ### Can an agent or a program use it? Yes, through API v1 or the MCP server (Claude, Cursor and others). See [API v1 and the MCP server](https://scholaris.joseluissaorin.com/en/knowledge/api-and-mcp.md) and the [page for agents](https://scholaris.joseluissaorin.com/en/agents.md). ### Who makes it? José Luis Saorín Ferrer, philologist and programmer, in Santa Cruz de Tenerife. Write to [jl@joseluissaorin.com](mailto:jl@joseluissaorin.com).