CONTENT CREATION

Extract and translate on-screen text before localizing

Runs OCR on a product screenshot via a Hugging Face model to read the embedded UI text, translates each string, and rebuilds the screenshot with localized copy in place.

CategoryContent Creation
Enginesim
Difficultyadvanced
Triggermanual
Steps5
Setup~25 min

How it runs

The automated pipeline, trigger to output.

  • TriggerOperator submits screenshot + target locale
  • ActionOCR the screenshot text via Hugging FaceHugging FaceHugging Face
  • LogicRoute low-confidence reads to manual review
  • ActionRegenerate screenshot with translated copyImage generation
  • OutputUpload localized image to R2 and return reportCloudflareCloudflare R2

What it does

Handles screenshots where the marketing copy is part of the image itself. It uses a Hugging Face OCR model to read the visible UI text, translates each detected string, and regenerates the screenshot with the localized text composited back into the same layout.

When to use it

When you only have a flat PNG export with no editable source layer and you still need accurate localized versions — for example reusing partner or legacy screenshots whose design files are gone.

How it works

  1. 1A manual trigger supplies the screenshot URL and the target locale.
  2. 2A Hugging Face OCR inference call returns the detected text regions and their bounding boxes.
  3. 3A logic step decides whether enough text was found to localize; low-confidence reads are routed to a manual-review flag instead.
  4. 4Each detected string is translated into the target locale.
  5. 5The screenshot is regenerated with the translated strings placed back into their original regions.
  6. 6The localized image is uploaded to R2 and the URL plus the extracted-text report are returned.

Set it up

What you configure once, before turning it on.

  1. 1
    Connect Hugging FaceModels, datasets, spaces — the open-source hub.
  2. 2
    Connect Image generationManaged Nano Banana image renders, metered per image.
  3. 3
    Connect Cloudflare R2Object storage, S3-compatible.
  4. 4
    Set each agent's modelWe leave models unset so you pick the tier — fast + cheap, or top-quality.
  5. 5
    Tune it to your dataEdit the prompts, filters, and field mappings so it matches how your team works.
  6. 6
    Test, then turn it onRun once against a sample, confirm the output, then enable the trigger.

Run this workflow in your colony.

14-day trial. No DevOps. No Sales call. Provisioned in under a minute.