Implement GPT like models that can enhance translation quality/consistency by being provided a context (eg extract from the image itself or from previous translations)
- Implement new translation models that support context (to be done in plugin)
- Implement as an option capability to turn context on and off (also for retro-compatibility with version where this is not a thing)
- Implement way to keep track of context
- Model should be agnostic of previous translations and just use the context provided during translation (be careful if model has memory)
- Context should probably be either a string or generalized to a database object that can be treated as an input to the pipeline to have consistent saving and re-fetching
Implement GPT like models that can enhance translation quality/consistency by being provided a context (eg extract from the image itself or from previous translations)