Could you reduce the 90-degree angle for a future release?. Please!
The 45-degree angle's fine, but the 90-degree angle's very inaccurate. Perhaps it would be better to reduce that view to 60 or 75 degrees. Anyway, thanks for this. I hope you can improve this even more!
The 45-degree angle's fine, but the 90-degree angle's very inaccurate. Perhaps it would be better to reduce that view to 60 or 75 degrees. Anyway, thanks for this. I hope you can improve this even more!
Perhaps we could wait until Z-Edit is released before training a stronger multi-angle LoRa model, as Z-image has very high performance. Furthermore, the 2509 model has many issues, and model 2512 might fix those problems, making LoRa less universal. Therefore, updating to a newer model now is not a good option.
Thanks, it's just my opinion. But I'd like you to keep it in mind for a new version since this is quite useful, although my experience is limited to 45 degrees and little more. Even the zoom outs aren't very precise, but I know that Qwen Edit 2509 is a complex model to train. Have you ever tried Kontext? It's also a great model, although it's a bit more difficult give an prompt, but you can use the same dataset. I'll be waiting for your future releases! 🫡
Thanks, it's just my opinion. But I'd like you to keep it in mind for a new version since this is quite useful, although my experience is limited to 45 degrees and little more. Even the zoom outs aren't very precise, but I know that Qwen Edit 2509 is a complex model to train. Have you ever tried Kontext? It's also a great model, although it's a bit more difficult give an prompt, but you can use the same dataset. I'll be waiting for your future releases! 🫡
I've tried Kontext on other datasets before and compared the two models. Clearly, Kontext doesn't perform as well as Qwen in most cases. It might be better in some aspects, but that would obviously take a lot of time to test. Therefore, the best approach is to wait for a more powerful model.