Earlier this month we explored turning centuries-old copperplate engravings from the Jode (the first picture Bible) and the Visscher Royal (a landmark set of 130 double-page engravings) into "living illustration" videos using Google's Veo video generation model to explore how the models reason about the rich layers of suspended motion depicted in these illustrations. Let's repeat those experiments with Google's latest and most powerful video model: Omni. Our earlier Veo experiments perfectly animated the original illustrations, preserving all of the original detail and simply making them "come alive", but yielded only fairly trivial sequences. In contrast, Omni yields far richer and interesting animations, but despite multiple attempts, fails to preserve the original illustration itself – instead using it merely as inspiration to create an "inspired by" animation, rather than animating the engraving itself.
Let's use the prompt we found to work best last time:
Turn this into a "living illustration" that completes the scene.
Let's start with the same image as last time:
Omni reframes the scene and alters things slightly, with an odd contortion and movement of the central person as they fall towards the water and even hallucinates a third arm for him:
And our second image:
This time, Omni entirely reworks the scene, changes the faces entirely and adds dialog, essentially just using the original image as visual inspiration, rather than attempting to animate the image as-is. Strangely it uses the same voice for both male characters and changes their faces dramatically:
And our Visscher Royal?
Again, Omni massively changes the scene, using the original image merely as inspiration, rather than attempting to maintain visual fidelity to it. At around the halfway point the ark's front changes dramatically:
Let's try this over again by asking Gemini 3.1 Pro to devise a prompt to use with Omni that would improve these results:
Give me a prompt to use with Omni to turn a historical engraving into a "living animation" that completes the scene. Animate the image as-is exactly.
Gemini gives us back this prompt:
Animate this historical engraving as a living, breathing scene while strictly preserving its original etched style, cross-hatching technique, and vintage aesthetic. Add subtle, realistic ambient motion to the environment and subjects: [e.g., clouds drifting slowly across the sky, water gently rippling, leaves swaying in a light breeze, and the figures breathing with slight shifts in their fabric]. The camera must remain perfectly still. Maintain the exact original color palette, line weight, shading, and composition. Do not alter, modernize, or morph the original subjects. The motion must be seamless and natural, bringing the scene to life exactly as it was originally drawn.
This time things begin ok, but once again Omni drifts off to recreating the scene from scratch as an "inspired by" video. In this one the woman walking with the front group mysteriously vanishes around the 7s mark:
Let's try being a bit more forceful with our instructions:
Turn this historical engraving into a "living animation". Do NOT alter the scene in any way. ONLY ANIMATE THE SOURCE IMAGE, DO NOT CHANGE IT.
Unfortunately, to no avail – and a strange hallucination in lower center with the deer:
