Post by Loiags in Manga Mode / Lens Mode does not overlay translations on the original text - itch.io

Thaluna: Real-Time Translator » Community

Viewing post in Manga Mode / Lens Mode does not overlay translations on the original text

← Return to tool community

Loiags4 days ago

Hi,

I tried the settings you recommended, but unfortunately the problem is still the same.

I used:

Lens Mode
Manga Mode: Enabled
Vertical Text: Disabled
OCR Language: English
Target Language: Japanese
Select Area: Only the comic page/panels

I reproduced the issue using the Diagnostic Build as well.

The Japanese translation boxes are still positioned away from the original English text. What I expected was for the Japanese translation to be placed directly over the original English text, as shown in the example images.

I have attached:

Original Comic Page
Incorrect Translation Overlay Position
Diagnostic Build console logs

I did not change the OCR language while reproducing the issue; it remained set to English.

Could you please investigate this as a possible overlay positioning/layout issue?

Thank you.

Diagnostic Build console logs

[Thaluna] Starting... HOME=C:\Users\ciosm

[Thaluna] Paddle models: C:\Users\ciosm\AppData\Local\Thaluna\.paddleocr

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\paddle\libs

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cudnn\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cuda_runtime\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cuda_nvrtc\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cublas\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cufft\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cusparse\bin

[CUDA_DLL_DEBUG] PATH prepended dirs=7

[TORCH_DLL_DEBUG] deferred torch dirs=1

[Thaluna] Models OK.

main.py:2588: DeprecationWarning: Enum value 'Qt::ApplicationAttribute.AA_EnableHighDpiScaling' is marked as deprecated, please check the documentation for more information.

main.py:2589: DeprecationWarning: Enum value 'Qt::ApplicationAttribute.AA_UseHighDpiPixmaps' is marked as deprecated, please check the documentation for more information.

[Thaluna] Starting... HOME=C:\Users\ciosm

[Thaluna] Paddle models: C:\Users\ciosm\AppData\Local\Thaluna\.paddleocr

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\paddle\libs

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cudnn\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cuda_runtime\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cuda_nvrtc\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cublas\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cufft\bin

[CUDA_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\nvidia\cusparse\bin

[TORCH_DLL_DEBUG] deferred torch dirs=1

qt.multimedia.ffmpeg: Using Qt multimedia with FFmpeg version 7.1.1 LGPL version 2.1 or later

[UI_DEBUG][translation_window.update_language_buttons] idx=8 engine=? model=English -> Japanese target='English' target_code='en' button='EN → JA' ocr_button='Japanese'

[STARTUP] Lazy OCR/model loading enabled; worker starts paused.

[LANG_RUNTIME][translation_window.apply_styles] idx=8 engine=? model='English -> Japanese' target_name='English' target_code='ja' font='Roboto' rtl=False

OCR language changed to: en

[UI_DEBUG][translation_window.update_language_buttons] idx=8 engine=? model=English -> Japanese target='English' target_code='en' button='EN → JA' ocr_button='English (Default)'

[DEBUG] Worker capture_active=False

[LENS_DEBUG][enter] main_idx=8 main_engine=None main_model=English -> Japanese worker_idx=8 worker_engine=? worker_model=English -> Japanese tw_idx=8 tw_engine=? tw_model=English -> Japanese ui_lang_button='EN → JA' main_ocr_device=auto worker_ocr_device=auto

[LENS_DEBUG][enter] target_visible=False (local model layout)

[LENS_DEBUG][snapshot_reader] lang=en persistent=False ocr_device_setting=auto use_gpu=True worker_idx=8 worker_engine=? worker_model=English -> Japanese

INFO: Could not find files for the given pattern(s).

paddle\utils\cpp_extension\extension_utils.py:711: UserWarning: No ccache found. Please be aware that recompiling all source files may be required. You can download and install ccache from: https://github.com/ccache/ccache/blob/master/doc/INSTALL.md

C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\paddleocr\ppocr\postprocess\rec_postprocess.py:1229: SyntaxWarning: invalid escape sequence '\W'

noletter = "[\W_^\d]"

C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\paddleocr\ppocr\postprocess\rec_postprocess.py:1455: SyntaxWarning: invalid escape sequence '\W'

noletter = "[\W_^\d]"

[2026/08/25 00:14:59] ppocr WARNING: The first GPU is used for inference by default, GPU ID: 0

[2026/08/25 00:15:01] ppocr WARNING: The first GPU is used for inference by default, GPU ID: 0

[2026/08/25 00:15:03] ppocr WARNING: The first GPU is used for inference by default, GPU ID: 0

[TORCH_DLL_DEBUG] preload: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\torch\lib\libiomp5md.dll

[TORCH_DLL_DEBUG] add_dll_directory: C:\soft\thaluna-real-time-translator-thaluna-preview-3.0-diagnostic\_internal\torch\lib

[TORCH_DLL_DEBUG] PATH prepended dirs=1

[LANG_RUNTIME][lens.refresh] mode=local engine=None model='English -> Japanese' target_name=None target_code='ja' font='Roboto' rtl=False source_name='English' source_code='en' groups=10

An error occurred executing the property metacall 2 on property "_language_button_glow" of TranslationWindow(0x2219cf911b0)

Traceback (most recent call last):

File "translation_window.py", line 371, in _set_language_button_glow

Thaluna.Dev4 days ago (1 edit)

Hi! Thank you for the screenshots.

I tested the same comic page and attached my result. The Japanese overlays are positioned over the original English text when the Lens toolbar is configured correctly.

My test used OpenRouter, but the translation provider does not control overlay positioning. The same Manga Mode layout rules apply to the built-in English → Japanese model.

I noticed that Manga Mode is not enabled in your attached screenshot. The checkbox directly next to “Manga Mode” is empty. This is separate from any Manga/OCR option found in the main settings.

Please use this exact workflow:

1. Enter Lens Mode.

2. Enable the checkbox directly next to “Manga Mode” on the Lens toolbar. Make sure a visible checkmark appears.

3. Keep Vertical Text disabled.

4. Set OCR Language to English.

5. Select English → Japanese or another provider with Japanese as the target.

6. Click Select Area and select only the comic page or relevant panels.

7. Do not zoom, resize, or move the page after selecting the area.

Please compare the toolbar in my attached screenshot with yours. In my screenshot, Manga Mode is visibly checked and Vertical Text is unchecked.

Please also check:

Windows Settings > System > Display > Scale

Your result may also indicate display scaling above 100%. Please tell me the exact value. As a diagnostic test, you can temporarily set it to 100%, restart Thaluna and YACReader, and repeat the workflow above.

If the problem remains, please send another screenshot showing Manga Mode visibly enabled together with the new overlay result and your Windows scaling value.

One additional note about the missing translations:

The Diagnostic logs show that you are using Thaluna’s lightweight built-in English → Japanese model. This is also why the Target Language selector is not visible on your Lens toolbar.

The built-in model is a basic offline option. With more complex comic dialogue, it may produce less accurate or mixed-language output, and in some cases it may return no translation at all. If the model returns an empty result, Thaluna does not create an overlay box for that text.

A missing box may therefore mean either:

- OCR did not detect the small or stylized comic text, or

- The built-in translation model returned an empty result.

My attached example used OpenRouter, which generally provides better and more complete Japanese translations. You can also use Ollama or a Custom OpenAI-compatible Local API.

The selected translation provider affects translation quality and whether usable text is returned, but it does not control the position of the overlay boxes.