Question 1

Getting started with nturn

Accepted Answer

nturn is a platform for translating documents with multiple AI models — Claude, GPT, Gemini and more. At its heart is the workbench, which brings translation memory (TM), glossaries, style guides and RAG together for each spreadsheet. ## Start in 3 steps 1. Sign in — sign in with your email or your Google account. ([Signing in](/help/login)) 2. Set up a project and file — pick a project in the workbench and upload the Excel, document or XLIFF file you want to translate. 3. Translate — select cells for a Flash translation, or use Precision translation to fine-tune quality and context. ([Flash vs. Precision translation](/help/quick-vs-detailed)) Supported file formats: xlsx, docx, XLIFF (standard XLIFF, memoQ mqxliff / mqxlz, Trados sdlxliff)

Question 2

What the workbench does

Accepted Answer

The workbench is a translation environment that brings source and target text, review columns, QA columns, translation memory (TM), glossaries, style guides and RAG into one grid (spreadsheet-style; plain documents and XLIFF are parsed automatically too). You edit source and target like a spreadsheet, with these features alongside: - Flash translation — select multiple source cells and translate them right away, including repeated sets that produce several alternative translations - Precision translation — choose columns and a start/end row, then set options and prompts in detail for large batch translations - Translation memory (TM) — reuse stored translations for consistency and speed ([more](/help/translation-memory)) - Glossary — keep chosen terms translated the same way every time ([more](/help/glossary)) - Style guide — apply tone and notation rules to translations ([more](/help/style-guide)) - Edit history — track who changed what, and when ([more](/help/edit-history)) - RAG search — pull relevant context from reference material in real time Use the left side panel to switch between sections — projects, workbooks, glossary, TM, Find patterns, RAG and more.

Question 3

Flash vs. Precision translation

Accepted Answer

The workbench offers two ways to translate. ## Flash translation Translates the cells you select immediately. It suits fast work with minimal setup — pick a domain, map the result column, and go. It also supports repeated-set translation, which produces several different translations for the same source cell. ## Precision translation Lets you specify quality mode, AI model, reference columns, surrounding-row context, glossary and style guide in detail. You can set domain, secondary type, tone and other options, edit the prompt to match your source text, and run automatic batch translation across a large row range.

Question 4

Working with translation memory (TM)

Accepted Answer

Translation memory (TM) stores sentences you have already translated and looks them up in real time when the same or a similar source appears, so you can reference or reuse them. ## Key features - Pre-translate — fills in sentences from your uploaded document that match the TM, reducing how much is left to translate. - Real-time saving — sentences you translate or edit accumulate in the TM automatically. - Match levels — source match (100%), source + context match (101/102%), fuzzy match, and partial match.

Question 5

Building a glossary automatically, and using it in translation

Accepted Answer

A glossary pins specific terms to a fixed translation. Use it for brand names, product names and technical terms that must not drift. ## How to use it 1. In the [Glossary management] menu, point at a file and let AI extract terms in bulk, then save them. 2. In the workbench, the Terms section of the side panel lets you look terms up, add source–target pairs by hand, and import or export. 3. Link a saved glossary during [Precision translation] and the terms are applied as you translate. Glossaries are managed per project, which keeps a whole team translating to the same rules.

Question 6

Creating a style guide and checking consistency

Accepted Answer

The workbench style features consist of three toolbar buttons (Generate · Check · Manage) and the "Style & review" section in the left side panel. The goal is to turn your project's translation style into a document, inject it into translations, and check consistency against it. ## 1. Generate a style guide — toolbar [Style | Generate] AI analyses the source ↔ target pairs you select in the sheet and produces a style guide in Markdown. - Specify the row range to analyse (start/end row; 30 rows or fewer is recommended), the domain · secondary type · style type, and the source/target languages. - Up to three reference columns can be layered in (Primary/Secondary/Tertiary). - Choose an AI model and a quality mode, and use the translation-pair preview to check before generating. - Copy · download the resulting guide, or save it to the database for reuse. ## 2. Check style consistency — toolbar [Style · Check] Checks translations against the selected guide and writes PASS / WARN / ERR verdicts and suggested fixes into the sheet. - Set the check range to the whole sheet or a manual row range. - Choose the 4 columns the results go into — verdict / suggested fix / detailed explanation / apply checkbox. - Pick from 7 check categories — lexical, morphological, syntactic, discourse, domain, Do's & Don'ts, and quick reference. ## 3. Manage guides — toolbar [Style · Manage] Manage saved guides and create derivatives. - Supports Save / Save as / Save to DB / Copy / Download / Delete. - Summarize (regenerate) — adjust the level of detail per component (exclude/basic/detailed) to regenerate the guide as a summary. - Create a translation — produce a translation of an English guide in another language so it is easier to follow. - Guides are classified automatically as regular / summary / translation. ## 4. The "Style & review" side panel - Active style guide — choose the guide to inject when translating this sheet. (Remembered per project and per open document.) - Find patterns / Scan patterns — generate a guide automatically from human edits, or scan unreviewed rows against the active guide. (See [pattern guide](/help/pattern-analysis-guide).) - Manage check results — work through results with Diff (compare the AI suggestion against the current translation or review, side by side), a cell colour toggle, and Apply all / Apply checked / Undo. You can switch between, restore and delete past check runs.

Question 7

Edit history and concordance search

Accepted Answer

Edit history tracks when, by whom (AI or human) and how each cell's translation changed, and lets you search and restore past translations. Press the [Edit history] button in the toolbar (shortcut Alt+H) to open a floating panel that does not cover the grid. ## Shared options - Follow the selected cell — the history switches automatically as you move the cursor to another cell. (Off by default.) - Show history across all versions — view history across several versions of the same file. - Scope — on sheets with a review column, choose "translation only" or "translation + review". ## History for this cell Shows changes to the one selected cell as cards, newest first. - The source and current target for that row appear at the top (copyable). - Each card carries: time, the kind of work (first translation / manual edit / style check / review), the edit type (AI suggestion · user edit · TM applied · reviewed, etc.), the AI model and provider, and a before → after comparison. - Undo restores the cell to the value at that point. Go to cell jumps to it. ## Full history for the open document & concordance search Audit every edit in the open document and search by value. - Type filters (translation work / style check / review / other) and date filters (today / this week / this month). - Search source or target — you can search the source text, not just translations. Use it like a concordance search to see how a term was translated in the past, across versions. - Group by cell — collapse several edits to the same cell together. - The footer shows statistics: cells edited, number of changes, AI/manual/reviewed ratio, number of editors, and average confidence.

Question 8

Importing and exporting Excel, DOCX and XLIFF

Accepted Answer

The workbench supports translation in Excel (XLSX), Word (DOCX) and XLIFF formats. ## The flow 1. Upload a document and it is extracted cell by cell or paragraph by paragraph into the grid. 2. Translate the source column or region, applying TM, glossary and style as needed. 3. When you are done, export to download an Excel / DOCX / XLIFF file with the original formatting intact. 4. Supported XLIFF formats: standard XLIFF, memoQ mqxliff and mqxlz, Trados sdlxliff DOCX and XLIFF files keep their heading structure, reference notes and other original formatting and conventions, and are saved as `originalname_translated`.

Question 9

Workflow role columns (source, target, review, QA)

Accepted Answer

The workbench goes beyond two source/target columns: you can assign several role columns for review and QA. Give each column a role and manage the translation → review → final workflow inside the sheet. ## Role types - Source — the original text to translate - Target — the first-pass translation, from AI or from you - Reviewed — a column a reviewer has corrected. You can assign several so reviewers work in parallel. - QA — the final quality-check column - Finalized — the authoritative column that merges the review results ## How to set it up 1. Open Row/column settings from the toolbar menu. 2. Assign a role to each column in the Workflow role columns section. (You can add more than one review column.) 3. Save, and progress is broken out into translation · review · approval stages. ## Merging into a final column To merge several review columns into one authoritative column, choose the source columns and the destination column, then press "Merge now". Rows where the values agree are written automatically; where they differ (a conflict) the cell is coloured so you can pick one yourself.

Question 10

Pattern-guide scans (automatic consistency checking)

Accepted Answer

Takes the style guide built by [Find patterns](/help/pattern-analysis-guide) and automatically scans unreviewed translations or review rows for rule violations. ## How to use it 1. In the workbench, open File > a specific version. 2. Run Find patterns to produce a pattern guide. There are two ways to give it something to analyse. - From edits you made in the workbench — if you edited the translation column directly, or wrote reviews in a review column, point Find patterns at that column. It analyses each cell's edit history (previous value vs. new value). - From columns the file already had — if the sheet arrived with source, translation and review all filled in, point Find patterns at the translation and review columns as soon as you open it. No editing first — it analyses the difference between the two columns. 3. If you have already run Find patterns, select the pattern guide under Active guide. 4. Run Scan patterns and unreviewed rows (target or review column) are compared against the pattern guide's rules. 5. Violations appear in the [Diff] window — it puts the AI suggestion next to your current translation or review — across 4 columns: verdict / suggested fix / detailed explanation / apply. 6. Review the suggestions, then apply the rows you checked in bulk, or undo them. A style guide derives linguistic rules — consistent tone, sentence structure, consistency elements across reference columns — from source and target pairs (well-translated sample segments a translator has polished). A pattern guide works differently: it compares a cell's edit history (previous translation vs. new translation) or the target against the review column, analysing exactly which sentence elements changed to derive linguistic rules. Because a pattern guide is grounded in the actual corrections made by human translators and reviewers, its consistency scans are more discerning.

Question 11

Building RAG (documents, sheets, TM and glossary)

Accepted Answer

RAG (retrieval-augmented generation) searches your project's reference material by meaning (semantically) and puts it to work in translation. In the RAG management menu you create a collection per project and build a RAG index over documents, Excel sheets, TM or glossaries, then query it in real time. ## Three ways to build RAG - Document RAG — add documents (text) such as manuals or terminology explanations. Documents are split into suitable pieces (chunking) and indexed. - Sheet RAG — add an Excel sheet. Specify the source, target and reference columns, and it is indexed by row and column so you can search source, target and reference material in real time with and/or conditions. - TM & glossary RAG — build RAG over the TM and glossaries stored in the database. This moves past the limits of keyword matching, letting you query TM and glossaries semantically in real time. ## Build order 1. Create a new collection in RAG management and choose an embedding model. Supported embedding models: OpenAI · Voyage AI (strong on multilingual, recommended) 2. Upload a document or a sheet, or select the project's TM and glossaries. 3. For plain documents, set the chunking method. (The default is recommended.) 4. Once indexing finishes, RAG matches appear in the context panel in real time as you move between cells. 5. Or use [RAG search](/help/rag-search) in the side panel to query [source, target, reference columns, TM, terms] in real time with and/or conditions.

Question 12

Settings for RAG search

Accepted Answer

Once a RAG collection is built, here is how to configure and use search — either from the RAG management screen in the top menu, or from workbench > side panel > Style & review > [RAG search]. ## How to use it 1. Choose a collection in [RAG management] or in [RAG search] under workbench > side panel > Style & review. 2. Enter a query and results come back ordered by similarity, limited to the number you set and to matches above your threshold. 3. When you use RAG search in the side panel, a sheet-type RAG object also shows the source, target and reference columns along with a badge telling you which field matched. 4. If a sheet is open in the workbench, you can jump straight to the cell a RAG match came from. ## Search modes - Semantic — finds matches with a similar meaning even when the wording differs. e.g. 보석 = Gem = 宝石 - Keyword — ordinary keyword matching, finding exactly matching words. - Hybrid — combines both: semantic matching, with a higher score the better the keyword match. To search from inside the workbench while translating, see [the RAG search side panel](/help/workbench-rag-panel).

Question 13

RAG search from the workbench side panel

Accepted Answer

In the workbench you can query collections semantically through [Style · review] > RAG search in the left side panel. ## How to use it 1. Open a document, sheet or XLIFF file in the workbench. 2. Click the RAG search button under [Style · review] in the left side panel. 3. Choose a collection that has a RAG index. 4. Set the search mode (semantic, hybrid, keyword), the match threshold and the number of matches. 5. For a sheet RAG, you can set the source, target and reference columns separately. 6. If the collection was built from TM and glossaries stored in the database, you can choose TM and terms as search targets. 7. Select a result and its source fills in automatically, or — if the same cell exists in the open document — you can jump straight to it. 8. Because the search is semantic, use the results as reference while you refine the translation in the open document. Search across a project or collection from the [RAG management](/help/rag-search) menu; search an individual collection from the workbench side panel.

Question 14

The real-time context panel (TM, term and RAG matches)

Accepted Answer

After you open a document, sheet or XLIFF file in the workbench, this panel automatically finds translation memory, terms and reference material related to the source and shows them in real time as you move between cells in the grid. You can see reference information while looking at the source. ## Matches shown - TM matches — finds similar sentences in the translation memory based on the source. Graded by how well they match: 102% (source and both neighbouring rows match exactly), 101% (source matches and one neighbouring row matches), 100% (source matches), fuzzy (character-based similarity across the whole sentence), and partial (part of a phrase or a run of words matches). - Term matches — pulls in every word in the sentence that matches a term registered in the glossary. - RAG matches — finds everything in the RAG collection above a similarity threshold whose meaning is close to the source. ## How to use it 1. Turn on Start context matching in the context panel. (Off by default.) 2. Toggle which items to show (TM · glossary · RAG). 3. Move between cells and the matches refresh automatically against that source. 4. Apply a TM match to the cell instantly with Ctrl+number, or a term with Alt+number.

Question 15

Signing up (email or Google)

Accepted Answer

nturn supports two ways to sign up. ## Sign up with email 1. Enter your email and a password on the sign-up page. 2. Your password must be at least 8 characters and combine upper- and lowercase letters, numbers and symbols. 3. Once you are signed up, sign in with that email and password. ## Sign up with Google The Continue with Google button handles sign-up and sign-in in one step. You can use nturn with Google authentication alone — no separate password to set. > Google accounts have their passwords managed by Google, so nturn does not offer password changes for them.

Question 16

Changing your password

Accepted Answer

Accounts created with an email address can change their password from Dashboard → the Settings tab. 1. Open Change password in the Settings tab. 2. Enter your current password and a new password (at least 8 characters). 3. For security you will be signed out and asked to sign in again. > Google accounts have their passwords managed by Google, so they cannot be changed in nturn.

Question 17

What is the Founder's Pricing lock-in?

Accepted Answer

[Payments are currently accepted only in KRW with cards issued in South Korea. We are adding payment methods for customers outside Korea.] As a thank-you to every Early Adopter who starts a paid subscription by 2026-12-31 (KST), we apply a 10% discount for life. Once applied it stays in effect for as long as you keep your subscription, and it comes off automatically at checkout — a monthly Studio subscription goes from ₩29,000 to ₩26,100.

Question 18

Can I change my plan at any time?

Accepted Answer

[Payments are currently accepted only in KRW with cards issued in South Korea. We are adding payment methods for customers outside Korea.] Upgrades take effect immediately; downgrades apply from your next billing cycle. Changing plans never removes or resets the credits you have already accumulated — they carry over.

Question 19

How does the [BYO API Key] discount work if I bring my own API key?

Accepted Answer

We do not support bringing your own API key yet. We will revisit BYO support once the service has stabilised.

Question 20

How are payments processed?

Accepted Answer

[Payments are currently accepted only in KRW with cards issued in South Korea. We are adding payment methods for customers outside Korea.] Payments are processed through Toss Payments, and credit cards are the only payment method. - Monthly and annual subscriptions — each payment grants the credits included in your plan. - One-off credit top-ups — independent of any subscription; buy credits whenever you need them.

Question 21

What is your refund policy?

Accepted Answer

- Within 7 days · unused: full refund - Within 7 days · partially used: refunded minus the credits you used (no cancellation fee) - After 7 days: refunded minus the greater of the pro-rated period or the credits used - One-off credit purchases: only the unused portion is refunded You can track token (credit) usage in real time on [your dashboard](/my/dashboard). Submit a 'Refund request' from [Billing history](/my/dashboard?tab=billing). Support will pro-rate the credits used since payment, remove the remaining credits, and refund to your original payment method.

Question 22

What happens when I run out of credits?

Accepted Answer

When you run out of credits, every feature that calls a frontier AI model is blocked. Features that do not call AI directly keep working — the workbench, edit history, building TM, adding terms, and the context panel. If you are low on credits you can upgrade your plan, buy a one-off credit top-up, redeem a promo code, request free bonus credits from the Billing tab, or check [free credits for joining Discord].

AI model settings

AI model

Temperature

Temperature

Temperature

Help

Frequently asked questions

Prompt templates

Edit template

AI model settings

AI model

Temperature

Temperature

Temperature

Frequently asked questions

Prompt templates

Edit template

Advanced prompt settings

Reference column prompts

Speaker Name

Tone / Emotion

Gender

Text Type

Structure Role

Formality

Character Limit

Context Reference

Notes

Reference ID