Skip to main content

Running translation

The pre-translate dialog​

Select one or more compatible project files and choose Pre-translate. The project dialog provides:

  • AI translation or TM only (no AI) mode.
  • A 50% to 101% TM threshold in TM-only mode.
  • Optional locking and confirmation of qualifying TM matches as part of PM file preparation.
  • TM matches, TB terms, cross-language matches, and available project research as AI context.
  • Empty segments only, Overwrite drafts, or Overwrite all handling.
  • Optional post-edit behavior that gives the AI the existing target as a draft when overwriting.

Pre-translate dialog

The dialog shows the primary account model configured under Settings > LLM Model Settings. Change the account setting there. Editor translation can also use a per-file model override without changing the organisation default.

Only workflow managers can use the project-page lock and confirm controls. Assigned linguists receive editor auto-translate only when Project Controls > Auto-translate is enabled for them.

Depending on the project's workflow type, translation may start automatically when processing finishes.

Choosing a model​

AI translation uses the account's configured provider and model, or an allowed per-file override. Models follow project instructions and can use TM/TB and research context. For content that should not reach a model, use TM only (no AI).

If a provider fails mid-run, the platform retries and can fall back to an alternative model rather than failing the job.

Privacy Mode​

For content that must not go to external AI providers, enable Privacy Mode in Settings > LLM Model Settings, with two Privacy Sources:

  • Platform: a hosted private model configured by LanguageOps.
  • Local: your organisation's compatible model server.

While Privacy Mode is active, a banner reminds everyone that translation is running through the private route.

Model settings​

The same settings page also controls:

  • Primary model and fallback models: provider order when the primary route fails.
  • Council Mode: multiple models translate and deliberate for selected high-stakes content.
  • Track Model Usage in History: record the model that produced a translation.

LLM model settings

What happens automatically​

Before any AI sees your content:

  • Exact TM matches are pre-applied: qualifying rows do not consume AI translation tokens.
  • Repetitions are handled once: repeated source content can be translated once and propagated.

During the AI run:

  • Termbase terms can guide terminology.
  • Fuzzy TM matches can provide reference translations.
  • Project guidance can provide selected research, style, and client context.

Translation modes​

When starting a run from the editor, choose a Translation mode:

  • Faithful translation: accurate and close to source.
  • Creative / transcreation: freer adaptation for marketing and brand copy.
  • Dialect / regional variant: a specified regional variety.
  • Monolingual polish: rewrite existing same-language text.
  • Amend existing translation: revise current targets using a supplied instruction.
  • TM only (no AI): apply qualifying memory matches and leave unmatched rows untouched.

An authorised user can override the model for this file without changing account defaults.

Document mode​

For content where segment-by-segment translation loses context, toggle Document mode for a whole-document pass. It supports faithful, creative, monolingual, amendment, and dialect variants, plus optional candidate-term extraction.

Prompt rules​

Projects can carry named Custom Instructions templates, such as rules about product names or formality. They are applied consistently to project AI runs.

Re-translating​

You can rerun translation on selected segments, for example after a termbase change. Empty segments only preserves existing targets, Overwrite drafts preserves approved work, and Overwrite all deliberately includes approved targets.

Segments covered by an active translation job are read-only until the job finishes or is cancelled. This prevents a background result from overwriting a simultaneous manual edit. Completed sessions must be unmarked before translation can run.

Monitoring progress​

Active jobs appear in a banner with live progress. Two details can make the displayed totals differ from the file total:

  • TM-prefilled and repetition rows can be excluded from the AI batch count.
  • Large runs are batched by word volume, so a batch may contain many short rows or a few large ones.

Expand the banner for per-stream progress. Project-page pre-translation is durable across refresh and navigation. When the run finishes, LanguageOps sends an in-app notification and an email when that preference is enabled.