Skip to main content

OCR Text Recognition and Translation

OVR Overlay Translator uses state-of-the-art locally accelerated OCR (Optical Character Recognition) technology to capture and extract text from the front of the VR environment's field of view.

Scope of Application and Recognition Capability

Unlike translation tools on the market that need to send the full screen to the cloud for processing resulting in extremely high latency, this software will, after the player controls the controller to delimit an area:

  1. Locally capture undistorted frame of the VR game immediately.
  2. Call the deployed high-performance inference library to process text region merging, which can well accommodate vertical, horizontal, and even irregular mixed line spacing game UI elements.

To learn how to enter fine-tuning paragraphs or sharpen blurry images, please refer to UI Settings -> OCR Settings instructions.

Translation Services

OVR Overlay Translator supports multiple translation services to meet different needs.

Translation Modes

Built-in Mode

Use the translation service provided by the software, ready to use out of the box.

Features:

  • No need to configure API keys
  • Requires network verification

Note: Some services may be temporarily unavailable and will appear grayed out and disabled in the interface.

Custom Mode

Use your own API keys, fully control the translation service.

Features:

  • Use your own account, no limits
  • Customizable models and prompts (LLM mode)
  • Requires manual configuration of API keys

Service Types

General Translation Engine

Traditional machine translation services with fast response speed.

VendorDescription
Tencent TranslatorTencent Cloud translation service
Microsoft TranslatorMicrosoft Translator
Volcano EngineByteDance translation service
Alibaba CloudAlibaba Cloud machine translation
Baidu TranslateBaidu Translate API
Youdao TranslateNetEase Youdao translation
Google TranslateGoogle Translate
DeepLDeepL translation service
Huawei TranslatorHuawei Cloud translation

LLM Translation Engine

Translation services based on large language models, with more natural translation results.

VendorDescription
OpenAIGPT series models
DeepSeekDeepSeek large models
Tongyi QianwenAlibaba Cloud Tongyi large model
DoubaoByteDance Doubao large model
SiliconFlowSiliconFlow model platform

Model Selection

When using the LLM translation engine, you can choose different models:

  • Auto: Automatically selected by the system default model
  • Specified Model: Manually enter or select the model name

In built-in mode, the system will automatically obtain a list of available models for selection.

Custom Prompts

When using LLM translation, you can customize translation prompts to optimize translation effects.

Default Prompt

You are a translation expert. The text to be translated is the result of OCR recognition, which may have some errors or truncated paragraphs or mixed languages. Synthesize all the text and give the most appropriate target language translation result.

Prompt Optimization Suggestions

  1. Specify translation style: Such as "use colloquial expressions" or "use formal written language"
  2. Handle technical terms: Such as "keep proper nouns in the game untranslated"
  3. Format requirements: Such as "maintain the original line break format"

Supported Languages

The software supports mutual translation between 30 languages:

LanguageCode
Auto Detectauto
Englishen
Chinese (Simplified)zh
Chinese (Traditional)zh-TW
Japaneseja
Koreanko
Frenchfr
Germande
Spanishes
Portuguesept
Italianit
Russianru
Arabicar
Hindihi
Thaith
Vietnamesevi
Indonesianid
Turkishtr
Greekel
Dutchnl
Polishpl
Swedishsv
Norwegianno
Danishda
Finnishfi
Hungarianhu
Czechcs
Bulgarianbg
Romanianro
Sloveniansl