Hey! No worries, Thaluna will run just fine on your RTX 5080 OC.
That warning does not mean anything is broken. It only means that the OCR and translation components bundled with Thaluna, PaddleOCR and Transformers, do not support GPU acceleration on RTX 5000-series cards yet.
For now, set these options inside Thaluna:
OCR Device: CPU Translation Device: CPU
CPU OCR is still perfectly usable. In typical use, it is usually only around 0.5 seconds slower than GPU OCR, depending on your CPU and the size of the capture region. Keeping the capture area limited to the text will also help.
The Translation Device setting only affects Thaluna's built-in offline models. It does not prevent external translation tools from using your GPU.
Your RTX 5080 is not wasted. You can use it through an external translation backend such as Ollama, LM Studio, llama.cpp, KoboldCpp, or another OpenAI-compatible Local API.
These tools manage the GPU independently. Thaluna can keep OCR Device and Translation Device set to CPU while the external backend uses your RTX 5080 for translation.
Ollama is usually the easiest option to configure. A model such as gemma3:4b is a solid starting point. You can also use OpenRouter if you prefer cloud translation.
Recommended setup:
OCR Device: CPU Translation Device: CPU Translation provider: Ollama or Local API External backend: GPU acceleration enabled for your RTX 5080
Select Ollama or Local API before starting OCR or Game Lens. Otherwise, Thaluna may unnecessarily load its lightweight built-in translation model on the CPU first.
Ollama setup video: https://www.youtube.com/watch?v=_9PjiXKfxUY
Local API setup guide: https://thaluna.app/local-api-guide
Translation model recommendations: https://thaluna.app/translation-quality-guide
RTX 5000 compatibility information: https://thaluna.app/hardware-compatibility
Long story short: your GPU is fine. Only Thaluna's bundled OCR and offline translation stack currently needs to stay on CPU. An external backend can still use your RTX 5080 normally and will usually provide better translation quality as well.