#7820 LLM纠错每批字幕行数: LLM大模型纠错时，每次发送多少条字幕，该值越大断句效果越好，一次性发送全部字幕最佳，但受限于最大输出token和上下文(max_ token)，过长输入可能导致超出AI限制而失败，默认20条字幕LLM纠错所用AI

The token / max_token is not filled in the "Advanced Options" page. The Advanced Options only lets you choose the batch size (LLM Correction errors How many subtitles are sent each time) and the AI provider (AI provider for LLM Correction errors). The actual API key and max token (max output tokens) must be configured in the settings window of the specific AI channel you selected for LLM Correction (default is DeepSeek).

Steps:

1. First, confirm / set which AI channel is used for LLM Correction:
   Menu -> Tools/Options -> Advanced Options -> General -> AI provider for LLM Correction errors (default: DeepSeek).
2. Then open that channel's own settings window for its API key and max token:
   Menu -> Translation Settings -> DeepSeek (or whichever channel you chose), and fill in the API key (SK) and the max output token (max tokens) there.

3. If you cannot find a "max tokens" field in that channel's window:

   • If the batch of subtitles is too large and causes errors such as Could not parse response content as the length limit was reached, simply lower the value LLM Correction errors How many subtitles are sent each time in Advanced Options (e.g. to 10–20), instead of increasing max tokens.
   • Note that some channels/versions may not expose a separate max-token input in the GUI.

Note: per the knowledge base, the API keys of cloud channels are entered in their respective settings windows under Translation Settings; the advanced options item only selects which channel to use, it does not contain a token field.

If you cannot locate the token field for your specific channel, please tell me the exact channel name and software version, and provide the latest log file from the logs/ folder.

Please refer to the related documentation:

• Advanced Options
• DeepSeek
• All Subtitle Translation Channels