Advanced translation
Run configuration.
The run configuration tells you exactly what the app will use before it spends time or money on OCR, translation, summary, or comparison work.
Fields
What each setting changes
Provider/modelThe AI service and model used for translation, OCR, summary, comparison, or fallback.
TemplateThe prompt style and output expectations for that task.
OCR modePlatform OCR, AI fallback, AI-only, or manual-only source extraction.
Output languageThe target language for translation and summary output.
Quality modeFast, Balanced, or Best quality image/input optimization.
ConcurrencyHow many pages may run in parallel during batch translation.
Provider decisions
Use the provider that fits the task.
GeminiRecommended first path for fast, cost-conscious online translation and broad beta testing.
Claude or OpenAIUseful for best-quality retranslation, comparison, editorial style, and alternate readings.
Mistral OCRUse for OCR/PDF extraction workflows when configured and available.
Local GemmaOffline Beta. Use when privacy/offline testing matters, but verify output carefully.
Quality
Fast, Balanced, or Best quality
- Fast sends smaller optimized images and may rely more on OCR-assisted input.
- Balanced is the default for most pages and whole-project work.
- Best quality sends larger image input and is better for difficult scans, dense pages, or publication review.
Fast does not mean a different translation standard. It means less input payload and faster provider work when the page is readable enough.
Preflight
What the app checks before a run
- Provider key or local model readiness.
- Model is available and appropriate for the selected task.
- OCR-only models are not selected for translation.
- Template kind matches the task: translation, OCR, or summary.
- Fallback provider is not identical to the primary provider.
- Large batch warnings show likely cost, time, and failure risk.