A voice you can trace.

Voxryn is a working source-backed conversation preview. It lets visitors import readable public website pages or paste text, find relevant passages, inspect citations, and listen with browser speech synthesis.

How sources work

Import up to eight sources. Each URL import reads one public HTTPS page using a server-side reader. It extracts readable text from the HTML returned by the host. It does not crawl a complete domain, run JavaScript, pass sign-in walls, or bypass bot protection. PDFs are not supported. Paste text when a page cannot be read.

Content is a snapshot at import time. Source cards and evidence show the import time. Refreshing a source requires removing and importing it again.

What answers mean

This studio uses extractive search in the browser with term-frequency ranking, synonym expansion, phrase boosts, and source diversity. It is not a generative AI model. Results are matching passages copied from the supplied sources. They are not independently verified facts, generated summaries, or proof that a source is trustworthy. Read the original context before acting.

Search can miss relevant material, especially synonyms, complex questions, and non-English text. If no passage matches, Voxryn declines to answer. It cannot detect every contradiction or determine whether information has become outdated.

Compare, brief, and inspect

Answer mode retrieves up to four matching passages while favoring source diversity. Compare mode shows one matching passage per selected source, and explicitly marks sources with no match. This does not establish whether sources agree, contradict each other, or are trustworthy.

Source Brief presents the opening readable passage from each selected source, or topic matches if you supply a question. It is an extractive orientation, not a full-document summary. Select up to eight sources, inspect their complete text, and find a phrase within a source.

TXT and Markdown imports accept 100 to 55,000 characters. Manual refresh re-reads a web source and reports whether its extracted text changed. Existing answers retain their previous quoted evidence. SHA-256 text fingerprints are available in secure browser contexts. They identify snapshots, not credibility.

Coverage maps and revisions

Coverage maps compare up to ten research topics against the selected sources. A matching cell opens an exact retrieved passage. No match means the keyword search found no relevant passage; it does not prove the topic is absent. The matrix is a snapshot at build time and must be rebuilt after changing sources. Export the map as CSV for review.

Manual refresh retains one previous text revision when the extracted text changes. You can also update text in the source inspector. A manually edited copy becomes a text source and loses its original website link, so revised text is not attributed to the website. Earlier answers and notebook quotes retain their original text.

Revision comparison lists added and removed paragraphs and counts unchanged ones. It ignores paragraph order and repeated identical paragraphs. A changed paragraph appears as removed plus added. Source revisions and research topic lists are included in saved workspaces. Follow-up suggestions come from frequent terms in selected sources, not generated reasoning.

Evidence notebook and collections

Pin up to 50 exact answer passages to a research notebook. Add up to 2,000 characters of your own notes per passage and export a cited Markdown brief. Notes are user-authored, not generated or verified by Voxryn. Notebook pins are stored in memory until you explicitly save or download a workspace; workspace restoration includes its notebook.

Collection import accepts public HTTPS page URLs, one per line, up to the available source slots. Each page is read sequentially and shows its own result. Stopping the queue takes effect after the current request. Failed pages are skipped. This does not crawl entire websites.

Workspaces and exports

Workspaces are saved only when you explicitly choose Save on this device. Up to six snapshots can be held in localStorage on this browser. They contain full source text, source selections, pinned passages, and notebook notes, not prior chat transcripts. Open a saved workspace to start a fresh conversation using its snapshots. Saving is not automatic. Remove stored copies from the Workspaces panel.

Download a portable workspace JSON file to restore sources elsewhere. Conversation exports are separate, available in Markdown or evidence JSON. Local session counters report only the current tab’s questions and retrieved passages; they are not traffic analytics.

Voice and embedding

The Listen button uses your browser’s text-to-speech service. Dictation uses SpeechRecognition where supported, and appears only when the browser provides that API. Browser vendors may process voice or text remotely under their own policies. You can always type.

The embed generator produces an iframe pointing to the same working studio in compact widget mode. A supplied public page is imported when the iframe opens. Choose a violet or mint accent and a widget height in the generator. Saved private workspace content is never included in the embed URL. Imports can fail if the original website blocks automated requests. Microphone permission requires browser approval.

Privacy

Pasted text and questions are searched in this browser tab. Imported URLs are sent to the Voxryn server, which fetches their publicly accessible page text. Sources and conversations remain in tab memory and are cleared on refresh unless you explicitly save a source workspace on this device. Saved source snapshots remain in localStorage until deleted; prior conversations are not saved there. There is no application server database or account system in this studio. Hosting providers and source website hosts may retain standard HTTP request logs.

No wallet connection, payment, or seed phrase is required. Avoid adding confidential information. Downloaded conversation files contain the questions and quoted source passages, so review them before sharing.

Token status

VOXR is a proposed ticker. No token contract address or Pons listing has been verified or connected to this website. There is no active token sale, staking, yield, fee sharing, token payment, or holder reward functionality.

The proposed development direction is usage-based indexing credits, paid publisher widgets, refresh services, publisher analytics, and eventually an agent API with metered payments. These are plans, not current capabilities. Ownership of a token would not guarantee returns.

Development direction

The next product stages would add permissioned publisher indexing, source freshness checks, a retrieval-assisted generative model with citation validation, and persistent publisher accounts. Payments and any token connection need their own implementation and verification before launch claims can be made.