Modify AI Translation Prompts (Prompt)
pyVideoTrans lets you customize the prompts used for AI translation and LLM re-segmentation to improve translation results. This article explains how to find and modify these prompt files.
I. Subtitle Translation Prompts
There are two different subtitle translation prompts. Which file is used depends on whether SendSRT is checked.
Prompt Used When SendSRT Is Checked
When the
SendSRTcheckbox is checked, this prompt is used automatically.
Prompt location: software directory/videotrans/prompts/srt/ folder
This folder contains prompt files for multiple providers. Each file is named after the provider:
| File name | Corresponding provider |
|---|---|
chatgpt.txt | OpenAI ChatGPT |
deepseek.txt | DeepSeek |
gemini.txt | Google Gemini |
zhipuai.txt | Zhipu AI |
xiaomi.txt | Xiaomi AI |
minimax.txt | MiniMax |
bailian.txt | Ali Bailian |
localllm.txt | Local LLM / Compatible AI |
This translation mode sends line numbers, timestamp lines, and subtitle text to the AI model as a valid SRT subtitle string, and expects the AI to return translated subtitles in a valid SRT format.
Editing note: Items in curly braces such as {lang} and {batch_input} are variables. They are automatically replaced by the program. Do not modify or delete them.
Per-Line Translation Prompt (When SendSRT Is Unchecked)
When
SendSRTis unchecked, this prompt is used.
Prompt location: under the software directory/videotrans/prompts/text/ directory, for example deepseek.txt
This folder also contains prompt files for each provider, with the same structure as the srt/ directory.
Editing note: Items in curly braces such as {lang} are variables. They are automatically replaced by the program. Do not modify or delete them.
II. LLM Correction Prompts
For speech recognition results, you can use an AI model to correct typos and other errors.
Prompt location: software directory/videotrans/prompts/resegment/llm.txt
You can switch this in Menu → Tools / Options → Advanced Options → General. By default, it uses OpenAI ChatGPT and compatible AI providers.
Use LLM for Resegmentation Instead of LLM Correction: Menu - Tools - Advanced Options - Speech Recognition Parameters - Use LLM for Resegmentation Instead of LLM Correction: Make sure the AI model used supports long context and long output. When checked, it will no longer only correct errors; it will re-segment the subtitles. The re-segmentation prompt is software directory/videotrans/prompts/resegment/llm_re.txt
IV. Quick Modification Guide
| Scenario | File to modify |
|---|---|
| AI translation results are not ideal (full subtitle mode) | The .txt file for the corresponding provider under videotrans/prompts/srt/ |
| AI translation results are not ideal (line-by-line translation mode) | The .txt file for the corresponding provider under videotrans/prompts/text/ |
| LLM correction is not effective | videotrans/prompts/resegment/llm.txt |
| LLM re-segmentation | videotrans/prompts/resegment/llm_re.txt |
Important reminder: Back up the original files before modifying prompts. Do not delete curly-brace variables such as
{lang}and{batch_input}; the program automatically replaces them with actual values.
Related Documents
- Improve AI subtitle translation quality — Translation mode comparison and glossary usage
- Better use of local LLMs as a translation provider — Local LLM configuration guide
- Modify AI translation prompts — Customize translation prompts
- Recommended settings for best video translation results — Optimal configuration for each stage
- Causes and solutions for "blank subtitle lines" after translation