CatScribe Docs

#Performance Guidelines

Translation speed depends on document size, engine choice, refinement settings, and your computer. Local and offline workflows are powerful, but they use your machine's CPU, GPU, memory, and storage.

#What Can Slow Down Translation

Large files take longer because CatScribe has more text to split, translate, review, and export. Complex formats can also require extra work to preserve structure.

Common causes of slower runs:

  • Very large PDF, EPUB, or DOCX files.
  • Long chapters translated as one batch.
  • Multiple AI refinement passes.
  • Large local models on limited RAM or GPU memory.
  • Heavy background apps competing for resources.
  • Scanned PDFs or files with complex layout.

#Practical File Size Guidance

These are workflow guidelines, not hard limits:

Project size Suggested approach
Short article or sample chapter Translate in one pass, then review.
Full chapter Use chapter-sized batches and review before continuing.
Full book Work chapter by chapter with a glossary and review passes.
Large PDF Test a small section first and check formatting before processing the full file.
Subtitle file Preserve timing and translate in sequence so context remains clear.

#Hardware Expectations

Local translation performance varies by hardware:

  • More RAM helps with larger files and larger local models.
  • A capable GPU can improve local model speed when supported by your setup.
  • CPU-only systems can still be useful, but large refinement passes may take longer.
  • Disk space matters when working with large projects, exports, and local model files.

If your machine feels overloaded, use smaller batches, close other apps, or choose a lighter translation profile for the first pass.

#Tips To Improve Performance

  • Start with a short test translation before running the whole document.
  • Translate by chapter, section, or scene instead of the entire book at once.
  • Use Auto Mode for a balanced setup when you are unsure which settings to choose.
  • Use lighter settings for the first draft, then refine only the sections that need it.
  • Avoid re-running the full project after every glossary change.
  • Keep source files clean and remove duplicated content before importing.

#Large PDF Expectations

PDF files can be slower and less predictable than EPUB or DOCX because PDF is often a visual layout format rather than a clean text structure. Text order, columns, footnotes, headers, and scanned pages can affect results.

For large PDFs:

  1. Translate a small sample first.
  2. Check whether paragraphs, headings, and tables are readable.
  3. If text extraction looks wrong, try using a cleaner source format when available.
  4. Split very large files into smaller parts before translating.