How to use the PixLate editor

The editor works in four steps.

1. 1
   Import — Upload one or more images, or a PDF. PDFs are split into pages automatically. When a PDF has selectable text, PixLate offers lower-cost Text Layer translation alongside Visual AI. Use Choose Files, File → Import Images/PDF, or drag files onto the Docs panel.
2. 2
   Select pages — Click a page in the Docs panel to select it. Hold Shift to select a range, Ctrl/Cmd to add individual pages. Only selected pages are processed.
3. 3
   Clean and/or translate — Choose one or more target languages in Translate. Leave source on auto-detect for most pages, or set it manually when you know the document language. If only part of the page should be processed, or an area should be preserved, draw Canvas regions before running Cleanup or Translate.
4. 4
   Review and export — Use the Layers panel to flip between original, cleaned, and translated versions. When satisfied, export selected pages as images or a PDF.

Templates pre-configure cleanup and translation settings in one click. They're a fast way to get sensible defaults for common content types without tuning each setting manually.

Each page keeps multiple layers stacked on top of the original. Layers are non-destructive — the original is always preserved.

Original

The source image exactly as imported. Always available, never modified.

Cleaned

The result of running Cleanup. Removes scan artifacts, improves contrast, or enhances color depending on your preset. Used as input for translation when "cleanup-first" is enabled.

Translated

The final output — source text replaced by translated text rendered in place. May be derived from the original (direct) or from a cleaned layer (cleanup-first).

Variants

Each time you run Cleanup or Translation, the result is saved as a new variant rather than overwriting the previous one. This means you can experiment freely — try a different preset, re-run with a style reference, or translate from a different source — without losing earlier results.

★ Default

The variant used for export and as the source for downstream translation. Each cleanup group and each language group has its own default. Change it by hovering a layer card and clicking the star.

Archive

Hides a variant from Compare mode and export without deleting it. Useful for keeping old attempts out of the way while preserving them for reference. Toggle with the eye icon on hover.

Duplicate

Creates a copy of a variant under a new ID. Use this to preserve the current result before re-running with different settings.

Cleanup improves image quality before translation. It's optional — if your source material is already clean and high-contrast, you can skip it and translate directly.

When to use cleanup

• Scanned physical books with dust, yellowing, or uneven contrast
• Phone photos of pages with shadows or distortion
• Low-resolution or faded source material
• When translation quality is poor on the raw source — a cleaned version often produces better OCR and translation results

Cleanup presets

Scanned Document

Removes scan artifacts (dust, specks, sensor noise). Improves contrast slightly. Preserves all content exactly — use this when you just want a cleaner scan.

Comics / Illustrated

Enhances color vibrancy and sharpens artwork. Preserves all text, borders, and panel dividers. Good for manga, comic books, and illustrated pages.

Faded / Old Document

Restores contrast, reduces yellowing and discoloration. Good for aged physical books or historical documents.

Photo Enhancement

Improves lighting, contrast, and color balance. Removes noise and sharpens detail. Use for phone-photographed pages.

Low Resolution

Upscales and sharpens detail, reduces pixelation artifacts. Use when source images are too small for clean OCR.

Custom

Write your own instruction for the cleanup model. Full control over what gets changed.

Cleanup styles

Styles control the color treatment applied consistently across pages — useful when processing a multi-page book to keep visual tone uniform.

Neutral

Removes color casts, normalises white balance. Good default for most content.

Preserve Original

Minimal color changes. Keeps the original appearance as close as possible.

Warm / Vintage

Adds a subtle warm or sepia tone. Suits older documents or a nostalgic aesthetic.

Cool / Modern

Crisp, cool tone. Suits contemporary manga or technical content.

High Contrast

Bold, enhanced contrast. Makes text and line art pop. Good for low-contrast scans.

Canvas regions let you control exactly which parts of the active page are sent to the model. Open them from the editor's left rail under Tools; on small screens, use the bottom Tools tab.

Drawing and editing regions

• Select Translate only or Protect, then choose rectangle or polygon drawing.
• Use rectangles for speech bubbles, captions, labels, and simple callouts.
• Use polygons for irregular text, angled labels, sound effects, or areas that wrap around artwork.
• Click an area in the Tools panel or on the canvas to select it. Selected regions can be moved, resized, reshaped, or deleted before processing.
• For polygon drawing, add points around the target area, undo the last point if needed, then finish once you have at least three points.

Custom translation for one area

A selected Translate only area can include custom target text. Use this when you already know the exact wording for a sign, title, name, caption, or sound effect. The custom text is sent with that area only, so it can be combined with other translate-only and protected regions on the same page.

Use Translate now for immediate processing and page-specific region protection. You can draw regions on each page, then run the checked pages together with Translate now. Check the result, including whether any QR code still scans, before exporting.

Background translation is currently available to selected accounts for English, High-quality output.

Background jobs target completion within 24 hours; delays can take longer. Finish uploading before closing the app. Results save as local editor layers when the project is open, connected, and writable; in-app alerts catch up when you return. Review results before exporting. Cloud results are retained for seven days after completion.

Background translation and Agent’s image workflow apply Protect and Translate only regions automatically. Protect wins where regions overlap. Regions are frozen when you submit a batch or prepare a run; later edits require a new job or run. Regular background translation uses originals; Agent’s image workflow preserves pixels from your selected visual base layer. Protected areas are masked in text and previous-page references too. Agent’s PDF text workflow does not apply canvas regions; use the image workflow for those pages.

See credit pricing

PDF translation method

PixLate recommends a translation method for each imported page. To use Text Layer mode:

1. Review the detected text boxes.
2. Choose the target language.
3. Translate the enabled text boxes or use the loaded-pages translation button.
4. Choose Preview → Apply to preview the current page.
5. Check the preview.
6. Choose Apply to Editor or Apply Page to apply the previewed page.

For multiple changed pages, use Apply All.

You can launch the Text Layer workspace with up to 50 selected pages at a time. This limit keeps browser memory predictable.

Translation presets

Literary

Preserves poetic devices, metaphors, and the author's voice. Best for novels, literary works, and story-driven content.

Academic / Technical

Formal language with precise terminology. Best for textbooks, academic papers, and technical diagrams.

Casual / Conversational

Natural, everyday language. Good for general-purpose translation where readability matters most.

Comics / Manga

Brief, punchy dialogue suited for speech bubbles. Preserves sound effects and exclamations. Best for comics and manga.

Religious / Sacred

Reverent, formal language appropriate for scripture and devotional text. Preserves proper nouns and divine names.

Custom

Write your own instruction for the translation model.

Refinement and reference images

In Visual AI, open Refinement to add instructions for wording, lettering, or layout alongside your style preset. These settings apply to the checked pages when you run translation; closing Refinement keeps them active. The Custom preset uses your instructions alone.

Choose Add references to select up to two cleaned or translated images from loaded pages. You can also select a cleaned or translated layer in Layers and choose Use for future translations. Click it again to remove that reference, or manage selections in Refinement. References guide lettering, tone, and text layout without copying artwork. Instructions and reference selections are saved with the project and reused until changed. Preview model inputs shows the source and reference images before you run.

Cleanup has its own style reference: select a cleaned layer in Layers and choose Use for future cleanup. Future AI cleanup runs match its color and contrast; batches run sequentially. Clear it in Cleanup to stop using it. This reference lasts until cleared or the editor reloads and is not saved with the project. Photographed-document cleanup uses deterministic processing and does not apply style references.

To improve one visual translation result, select its layer and choose Edit to change its pixels, or Retry to generate again from its source. More → Retry with this reference opens a retry with that result as a reference for that attempt; it does not change future-run references. Both create a new version and offer comparison. Regular background translation uses originals and does not apply sidebar styles, Refinement, or cleaned layers. Text Layer uses its own text controls.

Sequential vs Parallel

Compare mode splits the canvas vertically so you can view any two layers side by side. Useful for checking translation quality against the original, or reviewing how much cleanup changed a page.

• Enable from the View menu or press C
• Use the layer selectors in the toolbar to choose which layers appear on each side
• Drag the divider to change the split position
• Press C again or toggle in View to exit

Export collects layers from each selected page and downloads them as a PDF or image archive. You can export the active layer, the starred cleaned variant, or the starred translation variant for a selected language. Choose the winning variant in Review before exporting. Complete single-source PDF selections whose pages all have actionable Text Layer results can preserve selectable translated text; other combinations use the existing raster export path.

PDF import quality

When importing a PDF, this controls the resolution pages are rasterised at. Higher quality means better OCR accuracy but larger files and slower processing.

Low (150 dpi)

Fast processing. May miss small or dense text. Good for quick drafts and previews.

Medium (200 dpi)

Balanced. Good default for most modern printed books.

High (300 dpi)

Best accuracy for small text, fine detail, and dense layouts. Use for archival scans and technical content.

Find this reference in Help → Keyboard Shortcuts. Workspace shortcuts pause in text fields, other dialogs, and during input-method (IME) composition.

Main editor

PDF Text workspace

Zoom uses the same + / =, −, 0, and 1 keys. Use [ / ] to change pages. C, F, and Alt + 1–9 apply only to the main editor.

• To select a canvas text box, focus it and press Enter or Space.
• In Connect Text, select boxes in reading order. Each selection joins the pending passage in that order.
• To create a centered box without dragging, choose Add Text Box or Add OCR Box. Adjust its bounds under Position.
• In Edit view with the Select tool, use arrow keys to move the selected text box. Hold Shift to move it ten times the normal distance.

Docs

With the Docs list focused, ↑ / ↓ changes the active page and Space toggles its selection. Focused buttons and checkboxes keep their usual key actions.