The AI video generator Dream Machine from Lumalabs, which has been freely available for around six weeks, has been given a new function that allows short video clips with a specific target image (end frame) to be generated. Instead of a text prompt or an initial image (start frame), an image can be specified to which the model should “dream”. The clip length here is also 5 seconds.
Of the sample clips with which Lumalabs advertises the new function, we are particularly impressed by the one with the smoky lettering - with a bit of luck, this function could be used to create short typo animations in no time at all - fully automatically.
However, Dream Machine now also offers the option of defining both a source and a final image for video generation. Of course, this gives you even more control, but also a little more work and less chance. In an accompanying text prompt, this “keyframe interpolation” should also provide instructions on how the transition should be.
The videos, which are always 5 seconds long, are calculated by Dream Machine in a resolution of 1360x752. Incidentally, free access is currently limited to 10 generations per day; paying customers have priority.