Services
voix is a documentation studio in Chișinău, Moldova. Each service line below has a published price and its own page. They can be booked on their own or combined, for example a migration followed by AI-ready structuring.
7 service lines across 10 service pages, all priced in public.
We reply within 1 business day, from Chișinău, Moldova.
USD list prices, EUR on request. A fixed quote after a scoping call.
Any language pair: Romanian and Russian in-house, others via vetted partner linguists.
Service 01
Move a help center, wiki or pile of documents into Markdown in Git, on Docusaurus, MkDocs/Zensical or GitBook. From $4,500 (up to 150 pages), $9,500 (up to 500 pages), custom above.
Most SaaS help centers start in whatever tool support already had: Zendesk Guide, Confluence, a folder of Word files, a shared Google Drive. That works until you want docs reviewed like code, versioned with the product, or read by an AI assistant. Then the export button hands you markup, not content.
We convert the content to clean Markdown with frontmatter, rebuild the navigation, re-link images and attachments, and map every old URL to its new page so search rankings and support macros keep working. You review the result as pull requests, one slice at a time. The general case is documentation migration; platform detail is on Confluence to Docusaurus and Zendesk to Docusaurus.
Still choosing a target platform? The Docusaurus, MkDocs and GitBook comparison covers the trade-offs for help centers, and the migration checklist lists what to settle before anything moves.
Service 02
Turn a backlog of resolved Zendesk or Zoho Desk tickets into the KB and FAQ articles that should already exist. From $1,800 per 500 tickets.
Your ticket history is the most honest research you have: every question a customer asked because the docs did not answer it. Few teams ever read it as a whole, because nobody has time to go through a backlog ticket by ticket.
We work from an export of resolved Zendesk or Zoho Desk tickets with customer personal data removed. We group the tickets by the question behind them, rank the groups by how often they come up, write the missing articles, and rewrite the ones that exist but answer the wrong question. You also get a list of issues no article can fix, because they belong to the product. Support tickets to knowledge base has the full scope.
The tickets-to-knowledge-base method walks through the process step by step, if you want to try it in house first.
Service 03
Restructure docs so RAG assistants and support bots retrieve the right, current page instead of guessing. $2,500 audit; implementation from $6,000.
When a support bot gives a confident wrong answer, the model usually gets the blame. More often the problem is the content it retrieves: several versions of the same page, a critical caveat buried in a screenshot, a long article split mid-sentence by a chunker, and no metadata to say which product or version a page belongs to.
The audit reads your docs the way a retrieval pipeline does and returns a prioritized list of what causes wrong answers. Implementation fixes it: duplicates merged, headings that make sensible chunks, frontmatter metadata for product, version and audience, and plain-text versions of answers currently locked in images. Details are on AI-ready documentation.
Why your support bot hallucinates explains the failure patterns in more depth.
Service 04
Translation, MT post-editing and language quality assurance for docs and UI, in any language pair. $0.05 per word; LQA $45/hour.
We translate and review Romanian and Russian in-house. For every other language we work with vetted partner linguists who translate into their native language, and we stay your single point of contact for scope, terminology and delivery.
Where possible the work happens in your repository: Docusaurus i18n folders, MkDocs locale directories, or the string files behind your product UI. LQA checks accuracy, terminology, consistency with your glossary, and whether the translated page still renders correctly. Machine translation post-editing is available when the budget calls for it. Localization and LQA has the full scope.
Service 05
React and Docusaurus components, search, versioning and WCAG 2.2 AA remediation on the docs site itself. $75/hour or $560/day; accessibility audit from $1,900.
A docs site is a small web application, and it breaks like one: search that cannot find the page support links to, a versioning setup nobody dares touch, a custom component that traps keyboard users. We build and fix these in React and Docusaurus, billed by the hour or the day. See Docusaurus development.
The docs accessibility audit checks the site against WCAG 2.2 AA and returns a prioritized, developer-ready fix list, which is what you need when a customer asks about accessibility or ADA before signing. We can then do the remediation ourselves at the same rates. Common Docusaurus accessibility issues lists what usually comes up.
Service 06
Changelogs, release notes, versioned docs and screenshot refreshes, shipped with every release. Retainer from $900/month.
Release notes tend to get written last, by whoever is least busy, the evening before the release. Screenshots drift away from the UI one release at a time, and versioned docs fall behind until nobody trusts the version switcher.
On a monthly retainer we take that work off your team. For each release we read the merged pull requests and tickets, draft the notes and the changelog, refresh the screenshots and GIFs that changed, and cut the docs version. Nothing publishes until someone on your side approves it. See release notes and the release notes template and review gate guide.
Service 07
Human rating, evaluation and red-teaming of AI model output, in any language. $35/hour.
Models that answer in several languages need people who can judge the answers in those languages: whether they are correct, fluent, safe and faithful to the instructions. We rate, evaluate and red-team model output in Romanian, Russian and English, and bring in partner linguists for other languages.
The work follows your guidelines and runs in your tooling, billed by the hour. Multilingual AI evaluation has the details.
Scope
The ground every service line covers. A quote assembles the scope for your project from these; it rarely needs all of them.
Scope, price and delivery dates are fixed in a written quote before anything starts. Pricing lists every rate, and process sets out how a project runs from the scoping call to the handover.
Boundaries
A short list, published here so nobody discovers a boundary halfway through a project.
None of this is refused on principle. Out-of-scope requests are possible; under the terms of service they need written approval and a revised scope before they enter the plan, so the dates in your quote stay honest.
Free reading
If you want to understand the work before paying for it, or do it yourself, start here.
What we know about migrations, support content and AI-ready docs that fits in writing is published for free. Each guide covers one problem; the handbook walks through a docs-as-code migration end to end, from the audit to the launch.
More guides are on the blog. When you would rather have the work done than read about it, get in touch.
Timing
Delivery dates are fixed in the quote, slice by slice, after the scoping call.
We do not publish a standard duration, because the honest answer depends on page or ticket count, the source platform, the languages and how quickly your reviewers turn a slice around. The quote breaks the project into slices, each with its own delivery date and review gate. A slice ships as a pull request, and the next one starts once you have approved it or asked for changes. The process page explains how slices and review gates work.
Send the docs URL, the platform you are on and what is going wrong. We reply within 1 business day, and every project starts with a written, fixed USD quote.