As part of the user base that is specifically using TVAI for LQ to Low-MQ sources, I’d like to see the dev team refocus on training the AI to that end in Iris V3.
V1 was a game changer for that purpose, which was what was so exciting about it and the outlook of its future development. However, V2 seems more focused on high-MQ to HQ, which is fine, but it should be labeled as such and handled as a separate model rather than a V2 version.
I feel some of the feedback from users on subpar V2 performance vs V1 is due to our expectations on the vast improvement V1 brought to LQ and the added confusion of V2 still being labeled an LQ tool when it seems it was optimized for MQ+ sources. Therefore, Id like to make the suggestion to either keep V2 as a separate MQ-HQ model, similar to the Artemis models, and develop a seperate Iris model specifically for LQ to Low-MQ sources or retrain V3 to prioritize optimization of LQ so its more well rounded across all quality types and doesn’t leave the LQ user base under-served.
Keep up the great work Topaz!
