Improving the Quality of AI-Translated Subtitles

When using AI to translate SRT subtitle files, there are typically two approaches.

Method 1: Translate the complete subtitle block, including non-translatable elements like "line numbers" and "timestamp lines".

How to enable: In the software, select Send Complete Subtitle Block.

Example: The complete block with formatting is sent.

 1
 00:00:01,950 --> 00:00:04,950
 Organic molecules have been discovered in the Five Old Star System.

 2
 00:00:04,950 --> 00:00:07,902
 How far are we from a third-type encounter?

 3
 00:00:07,902 --> 00:00:11,958
 The microwave imaging mission has been ongoing for a year.

Advantage: Considers context, resulting in better translation quality.

Disadvantage: Besides wasting tokens, it may also cause formatting errors in the translated subtitles, making the result an invalid SRT format. For example, English punctuation like , : might be incorrectly changed to Chinese punctuation, or line numbers and timestamps might be merged into a single line, etc.

Method 2: Send only the subtitle text content, then replace the corresponding text in the original subtitle file with the translation.

How to enable: In the software, deselect Send Complete Subtitle Block.

Example: Only the subtitle text is sent.

 Organic molecules have been discovered in the Five Old Star System.
 How far are we from a third-type encounter?
 The microwave imaging mission has been ongoing for a year.

Advantage: Guarantees the translation result will be a valid SRT subtitle format.

Disadvantage: It's also quite obvious. Translating line by line cannot consider context, significantly reducing translation quality.

To address this issue, the software supports translating multiple lines at once (default is 15 lines), which can provide some context.

However, this introduces a new problem: Different languages have different grammar rules and sentence structures. It's possible that 15 lines of original text become 14 or 13 lines after translation, especially when the first and last lines of the block are part of the same grammatical sentence.

When the 15-line original subtitle block no longer has 15 lines after translation, it will certainly cause subtitle chaos. To solve this, if the number of lines in the translation result does not match the original, the software will revert to translating line by line, ensuring the line count remains exactly the same, sacrificing context consideration.

---

Note: The problems mentioned in the first method may occur, resulting in an invalid SRT subtitle format. This can lead to parsing errors or loss of all content after the error. It is recommended to use this method only with sufficiently intelligent models, such as GPT-4o-mini or larger models. For locally deployed models, this method is not recommended. Due to hardware limitations, locally deployed models are generally smaller and less intelligent, making them more prone to formatting errors in translation results.

Enabling the First Translation Method:

Menu -> Tools/Options -> Advanced Options -> Subtitle Translation Section -> Send Complete Subtitle Block for AI Translation

Adding a Glossary

You can add your own glossary to any prompt, similar to the example below.

**During translation, you must use** the glossary I provide for translating terms to maintain consistency. The specific glossary is as follows:

   * Transformer -> Transformer
   * Token -> Token
   * LLM/Large Language Model -> Large Language Model
   * Generative AI -> Generative AI
   * One Health -> One Health
   * Radiomics -> Radiomics
   * OHHLEP -> OHHLEP
   * STEM -> STEM
   * SHAPE -> SHAPE
   * Single-cell transcriptomics -> Single-cell Transcriptomics
   * Spatial transcriptomics -> Spatial Transcriptomics