Please provide more in-depth documentation for Full-Reference Mode
The prompt guide for the T2VA / I2VA / FL2VA / L2VA modes is sufficient, but the guide for the reference mode only covers some of the most basic use cases, and only hints at other, advanced capabilities. The community has been scrambling for the past few weeks trying to piece together the correct prompt structure for many of those more advanced cases, mostly with little success. The prompt structure for the reference mode is simply too complex and the current guide provides too little information for us to figure it out fully through experimentation and trial&error alone.
Please, I beg you, provide a more comprehensive documentation for the Full-Reference Mode, or if you don't have the time, at least consider publishing some of the captions you used during training - even without the actual training data - so we have SOMETHING to guide us in our explorations of this fantastic model you've given us.
The most powerful tool is useless without the knowledge to wield it - I feel like someone handed a pilot’s cockpit and taught me only how to start the engines.
Thank you for this fantastic model. I hope you will consider my request.