JDM Translation Tips: Getting Better Decipher Results

The quality of your auction sheet decipher output depends significantly on the quality of the input image. These tips help you get consistently accurate translations.

Image quality requirements

The decipher performs best on images that are: at least 1200px wide, well-lit without shadows across the text, not rotated more than 5 degrees, and scanned at 150 DPI or higher if sourced from a physical document.

Most auction house digital sheets (TAA, USS online) produce good quality images automatically. Physical sheets photographed with a smartphone are the most variable — ensure good lighting and hold the camera directly above the sheet.

What the AI reads well

Standard USS, TAA, and JAA inspection forms from 2010 onwards are well-represented in the training data. The AI reliably extracts: overall grade, interior grade, mileage, all standard damage codes and their locations, odometer status notes, and standard inspector comment abbreviations.

What the AI reads less reliably

Handwritten additions outside the standard form structure are less reliable. Very old forms (pre-2005) use different layouts that may cause field misalignment. Forms from regional auction houses (local cooperatives, agricultural equipment dealers) use non-standard formats.

Improving results with context

The decipher sends listing context (make, model, year) along with the image. For vehicles with unusual configurations or rare models, adding detailed listing information helps the AI interpret ambiguous codes correctly.

Running a second decipher

If you received a result that seems inaccurate, you can run a second decipher — this consumes another quota credit. Consider reviewing the original image first to check scan quality before spending another credit.

Language-specific nuances

For Arabic output, right-to-left text rendering in the app is handled automatically. For Japanese output, the decipher will preserve original Japanese terms for concepts with no direct translation (specific damage code abbreviations, for example) alongside their English meaning.