#8515 so the prblem i am facing is that the video contains two languages mostly chinese and at some point japanese so the tran

This happens because the Spoken language is set to a single language, so the recognizer forces every segment into that language and the Japanese parts get "translated" into Chinese characters/sounds.

Fix options, from simplest to most reliable:

1. Use Auto Detect + a multilingual Whisper model

1. Set Spoken to Auto Detect (bottom of the drop-down).
2. STT channel must be faster-whisper(Built-in), openai-whisper(Built-in) or Qwen-ASR(Built-in) — other channels require a fixed language and will error or mis-recognize. Use model large-v3 (or large-v3-turbo if VRAM is limited).
3. Do not embed bilingual subtitles (Hard/Soft Subtitles (Bilingual)) when Auto Detect is used.
4. For the translation channel use an AI model (DeepSeek / ChatGPT / Gemini), because free channels such as Google/Microsoft return the original text untranslated when the input mixes languages.
   Note: Whisper detects language per audio window, so short Japanese interjections inside long Chinese speech may still be missed.

2. Split the video at the language boundary (most accurate for mixed content)

1. Use Menu – Tools – Trim Video by Subtitles (or any cutter) to cut the video into a Chinese part and a Japanese part.
2. Apply the numbering/import structure described in the local-SRT import doc, or translate each part separately with Spoken set correctly per part (Chinese → Qwen-ASR/FunASR, Japanese → openai-whisper large-v3).
3. Merge the results with Menu – Tools – Merge Video & Audio / Combine Two Subtitles.

3. Improve the recognition input

1. In More Settings, check Denoise, or Separate Vocal/BGM if there is background music.
2. Check LLM Correction (Advanced Options → Speech Recognition → LLM Correction Channel, DeepSeek by default) to fix typos and punctuation.
3. In single-video mode, use the subtitle editing window (Stage 1) to correct the Japanese lines manually before translation.

Please refer to the related documentation:

• All speech recognition channels
• Improving segmentation results
• Importing local subtitles or vocals