Subtl Sign in

Lexicon-governed translation

Your terminology,
every language,
every medium.

Subtl builds a living lexicon from your work. Annotate terms, translate with precise context, and keep terminology consistent across subtitles, documents, presentations, and ebooks.

Generic translation tools dump everything into one prompt.

Subtl sends only the terms that matter. A 500-entry lexicon produces 5–15 context entries per batch — precise enough for the LLM to get it right, lean enough to keep costs down.

Living Lexicon

Your lexicon evolves through use. Annotation discovers terms, translation reveals gaps, review refines definitions. Reuse it across every project.

Annotation-Driven

Mark up meaning, disambiguation, and context in the source text. LLM assists with matching; humans decide what's correct.

Per-Language Governance

Each language is governed independently. Same term, different register — a subtitle reads differently from a legal document.

Three phases. One iterative loop.

Translate first, spot errors, go back to annotate better, then re-translate. The loop is the workflow, not an edge case.

1

Source & Annotate

Upload your content. Annotate terms — link to lexicon entries or propose new ones. Auto-match catches what you miss.

2

Translate & Review

LLM translates with your lexicon as context. Review per block, approve or revise. Bad output? Go back to step 1.

3

Export

Export in the original format — timecodes reattached, slide structure preserved, document formatting intact.

Every medium, one platform.

Import your source content. Subtl understands the structure and preserves it through translation.

Subtitles

SRT / VTT

Documents

DOCX

Presentations

PPTX

Ebooks

EPUB

Web

HTML

Plain text

MD / TXT

The lexicon is the product.

Most translation tools treat glossaries as an afterthought. In Subtl, the lexicon is the center of everything — a governed knowledge base that grows smarter with every project.

Reusable across projects. Build once, apply everywhere. Fork for variations.

Context, not pollution. Only matched entries reach the LLM. Focused prompts, better output.

Human-governed, LLM-assisted. The LLM proposes; your team confirms. Every translation is traceable.

Team workflows built in. Solo, managed, or self-serve — with or without review gates.

Stop translating blind.

Build your lexicon, annotate with intent, and let every translation carry the right context.