Are I2V & T2V workflows actually different?
If both workflows have the "Switch to Text to Video" Boolean toggle, why do we need 2 separate workflows? Is there a difference under the hood that's not obvious? I've used one work flow both ways, and except for the odd text prompt injecting an old prompt into the first generation (mentioned here in another thread), it seems to work fine.
I dont know if they have actual differences but i can imagine running deviating sigma's and schedulers can produce very different results for t2v/i2v
There is no under the hood difference. The only difference between 2 flows is that I2V flow inlines the provided image as the first latent frame and freezes it while T2V starts with an empty latent frame.
Other than that the flows are identical. In the official LTX workflows templates t2v and i2v are combined in a single flow with "use image" boolean switch.