Frequently asked
questions.

Because we quote directly and put it in writing. Lisan sells across several markets with different deployment models (cloud, self-hosted, on-premises), and a one-size price would be wrong in both directions. The free tiers are genuinely free; paid plans are quoted per organization.

No. Every flagship has a working free tier: free desktop apps, free-to-start web workspaces. Trials with expiry dates exist on some products' premium plans, and their sites say so explicitly.

If your team adopts more than one Lisan product, the platform price beats the sum of the parts. Talk to us about which combination fits.

No. When you buy a Lisan product, its modules come along. Vquity includes every equity module on one plan; Vinance includes its ledgers, payroll, and inventory without add-on fees.

The way your procurement requires: deployment-scoped quotes, security review support, and contracts with your entity. More than 24 government entities have been through this with us. Start the conversation.

NoteX: a Notion-style REST API with scoped keys, plus an MCP server with ten tools. It is live for NoteX web workspaces; reference lives on notex.lisan.org. More surfaces open up as they mature.

Keys are scoped read-only or read-write, optionally restricted to a page subtree, optionally expiring, with usage logged. You create and revoke them yourself.

No. Every flagship works standalone, and many customers start with exactly one product. The platform advantages (one account, shared data, one admin story) arrive when you add a second.

The flagships are the deep, named products (NoteX, WriteX, TranslateX, MeetriX, Vinance, Vquity). The workspace is the connected suite of everyday apps around them. Same account, same design. See the workspace.

In-house models carry the Arabic-critical paths like transcription, proofreading, and translation. For assistant features, products integrate leading commercial models, and desktop add-ons let you bring your own AI accounts.

Yes. NoteX exposes a Notion-style REST API with scoped keys and an MCP server, so any MCP client can search, read, and write your knowledge base under your control. See NoteX.

No. The pattern across the platform is propose-then-approve: AI drafts, suggests, and extracts, and a human clicks before anything is applied.

Yes. Every product is fully English-capable, and several (NoteX, Vinance, Vquity) compete on engineering merits that have nothing to do with language. Language depth is about being strong where others are shallow, not exclusivity.

MeetriX is built for the way the region actually speaks in meetings, including dialects and Arabic-English code-switching. For specifics on your dialect mix, talk to us and we will show it on your own calls.

Yes. On-premises deployment is how Lisan serves government customers whose data cannot leave their infrastructure. Talk to us to scope one.

You export it, in open formats. Products that store notes and documents as files make that trivial by construction; the rest provide full export paths. Exit rights are part of the design, not a negotiation.

Lisan does, end to end: installers are code-signed and served from updates.lisan.org, Lisan's own release infrastructure, with stable and beta channels.

Lisan (LISAN AI - FZE) works from five cities: Dubai, United Arab Emirates; Riyadh, Saudi Arabia; Cairo, Egypt; Warsaw, Poland; and Madrid, Spain.

More than 24 government entities work with Lisan, alongside corporate customers across the region. Products like WriteX also serve institutions from ministries to newsrooms.

We are always interested in exceptional people who care about language and AI. See careers for what we look for and how to reach us.

Because in language checking an unnecessary change usually costs more than a missed one: it has to be reviewed, rejected and sometimes reverted by hand. F0.5 weights precision above recall to reflect that. We publish F1, precision and recall as well, and L3 leads on all four, so the choice of headline metric does not change the ranking at the top.

No. Lisan Research designed and ran it. What we can offer instead is the full method: the verbatim prompt, the single-output rule, the exact-span scoring definitions, the normalization steps and every rate in the appendices. Anyone with a linguist-reviewed Arabic set can repeat the protocol, and the WriteX editor is free for that purpose.

Several reasons show up in the rows. The scoring is strict: a model earns credit only when it changes exactly the span the linguist changed, and a punctuation edit can be its own span. The prompt forbade rephrasing and diacritics, so editorial-standard rows and classical rules such as hamza seats inside common technical terms were often left untouched. And the profiles differ: Claude Fable 5.1 changed little and was usually right, while GPT-5.6 Sol changed a great deal and was usually not. Recall of 20.15% and 41.08% are the two ends of that range.

No. That rate is conditional: of the spans L3 located correctly, 96.38% carried the right replacement after Arabic normalization. Its recall was 66.85%, and its strict whole-sentence match was 55.96%. Read the three together.

L3's figure is a direct measurement on an NVIDIA A10 with a three-times full-stack factor applied. The large language model figures are cited anchors: Google's Gemini Apps disclosure and the IFP School per-query estimates, scaled to 1,000 tokens. They are not measurements of the tested API models, and we show the two-times to four-times sensitivity range so the reader can see how much the factor matters.

CER, the character error rate, is the share of characters in a transcript that had to be added, removed or changed to match what a human verified was said. This report scores CER after the normalisation rules listed under "How we measured", so dialect and standard spellings, Latin and phonetic English, and numeric forms count as equivalent. Word error rate is not reported here, and earlier MeetriX figures published as WER are superseded by this evaluation.

No. Lisan Research ran the evaluation. The protocol, scoring rules, examples and evidence map are published on this page so the scoring can be reproduced, and a pilot on your own meetings with a human-verified reference is the way to check the result on your material.

Close. It recorded 2.26% against M3's 1.62%, a gap of 0.64 percentage points, and the review credits it with strong handling of English terms and low content loss. The differences that remain are Latin versus phonetic spellings of terms, the concentration of artifacts, and the speaker-linked workflow around M3, which was outside the transcription comparison.

In the evaluated deployment the transcription path called no external transcription service. M3 can run within customer-controlled infrastructure, keeping meeting audio, transcripts and summaries inside the organisation's data-governance boundary.

Yes. A pilot runs M3 on a sample of your meetings against a human-verified reference, with the scoring rules shared before the run. Write to support@lisan.com or book a pilot through MeetriX.

They are written from production systems and prepared for submission to peer-reviewed venues. The engineering they describe is running in products today; the publication process is underway.

Yes, it is published. The M3 evaluation report states the protocol (same audio, same human-verified reference, same rules for all seven systems), the full scoring rules for CER after normalisation, the configurations of every compared system, five transcript examples, an evidence map with the status of every claim, and the limitations. Read it at the M3 report. For a pilot on your own meetings, talk to us.

Sales. It is the widest door, and anything that belongs somewhere else gets walked over internally instead of bounced back to you.

Sales, support, and press inquiries sent on a working day usually get a human reply within one business day. Investment and partnerships take up to two, and careers up to five. Deployment conversations get a call scheduled, not a ticket number.

It is the fastest way to pick a conversation back up. Quote it in any later email, from any address, and we can find the original message, who has it, and what was said.

Yes. The Arabic contact page carries the same desks and the same form, and the confirmation comes back in Arabic. Either way, someone who reads your language answers.

It is used to answer you and to keep a record of the conversation, and it is stored on Lisan infrastructure. The details are in the privacy policy.

Yes, and you should: transcription on your meetings, proofreading on your documents, translation on your files. Demos on our sample data prove less than demos on yours.

Yes. Lisan Hub and every desktop app are free to download and install, and each product has a real free tier. Paid plans only come in when your team needs more.

Every installer is code-signed by Lisan and served from our own release infrastructure at updates.lisan.org, not a third-party CDN. Lisan Hub also verifies the integrity of every download before installing it.

Lisan Hub quietly keeps every installed app on the latest release in the background. No update pop-ups, no downloading new installers by hand.

Windows is available today; macOS and Linux builds are planned. In the meantime, every Lisan product runs fully in the browser on any platform from your same account.

Yes. Signed installers, silent-install flags, and a single versioned release feed make Lisan desktop apps manageable across a fleet. Talk to us about organization-wide rollouts.

Lisan is a deep AI technology company working from Dubai, Riyadh, Cairo, Warsaw, and Madrid, building an AI-native productivity platform for government and corporate organizations. Born in the MENA region and built for anywhere, the platform spans six flagship products and a connected workspace of more than 20 apps.

Six flagships: NoteX (knowledge platform), WriteX (Arabic-first proofreading), TranslateX (document translation), MeetriX (AI meeting notes), Vinance (finance), and Vquity (equity management), plus the Lisan workspace with docs, sheets, slides, tasks, CRM, and more. See all products.

No. Everything works fully in English, TranslateX alone covers 20+ languages, and WriteX checks English and Spanish alongside Arabic. We proved the platform on Arabic because it is one of the hardest languages for software to get right; the same engineering carries every language we add.

Yes. Sovereignty is core to how Lisan sells to government: self-hosted servers and on-premises deployments are available for organizations that need data to stay inside their infrastructure. See sovereign deployment.

Each product is free to start: open the web app or download the desktop apps from Lisan Hub. For teams and organizations, talk to us and we will scope what you need.

Yes. On-premises deployment is how Lisan already serves government customers, and self-hosted server options cover the middle ground. See sovereign deployment.

Write to support@lisan.com or use the contact page. We will scope requirements, demonstrate on your own documents, and provide what your review needs.

No. Start with one product (most teams start with MeetriX or Vinance), and adopt siblings when ready. Vinance imports from Xero, QuickBooks Online, and Zoho Books.

Every product has a free tier; team plans are priced per product, and bundles reward adopting more of the platform. See pricing for how it works.

Yes. The free tiers are real working versions, not trials with a countdown. You will know if you need more before we ask you to pay for it.

Always. Open export formats across the platform, and NoteX keeps notes as plain files you can open with anything.