Image-to-Video: how to stop i2v from 'releasing' your start frame into a completely different scene

by @jordansantos_films in c/seedance-community · · Image-to-Video

The classic i2v failure: you feed a gorgeous start frame, the first second is perfect, and then the model 'releases' and drifts into a related-but-different scene — the room changes, the character restyles, the color shifts. 2.5 reduced this a lot but it still happens on ambitious clips. Here's how I keep the start frame honoring itself.

LOW motion intent. The more motion you request, the more license the model takes to reinvent. If you prompt 'she turns her head slowly,' it stays home. If you prompt 'dynamic action sequence,' it treats your start frame as a loose suggestion and leaves. Motion magnitude = drift license.

DESCRIBE the start frame in the prompt. Don't just feed the image and prompt the ACTION — also describe the SCENE that's already in the image. Redundant? Yes. But it anchors the model: the image says 'red kitchen' and the prompt ALSO says 'red kitchen,' so it has two votes to stay put instead of one.

Short clips. Drift compounds over time. A 4-second clip drifts a quarter as much as a 16-second one. Generate short, chain if you need length.

The mental model: the start frame is a LEASH, and motion magnitude is how much slack you give. Ask for calm motion and a described scene and the leash stays short.

349 points · 6 comments

Comments

  • @oscargarcialab4675 · · 67 points

    'Motion magnitude = drift license' is the cleanest way anyone's put the core i2v tradeoff. More motion literally means more permission to reinvent, because high motion REQUIRES change and the model can't tell 'change the pose' from 'change the scene.' Ask for calm, get fidelity. Ask for chaos, get a new world.

  • @dianajohnsoncreative · · 44 points

    It's a fundamental tension, not a bug — motion IS change, and the model's concept of 'change' is holistic, not surgical. It doesn't know you only wanted the HAND to move. So you keep total requested change low to keep UNWANTED change low. Calm prompts = loyal clips.

  • @rileywright_digital775 · · 41 points

    Describing the start frame redundantly in the prompt is the trick I didn't know I needed. 'Two votes to stay put' is such a good way to think about it. The image AND the text both insisting on the same scene is way stronger anchoring than either alone. Doing this from now on.

  • @zarasilva_studio · · 35 points

    Would add: match your start frame's IMPLIED motion. If the start frame is a static portrait, don't prompt a car chase — there's no momentum in the image to continue, so the model has to INVENT all of it, and invention = drift. Feed a frame that already looks mid-motion if you want motion. The image should 'want' to move the way you're asking.

  • @omaryoung_digital461 · · 27 points

    The start frame should already imply the motion you want — a frame caught mid-stride continues into a run naturally, a static standing pose fights it. Choosing a start frame with the right implied energy is half the i2v battle and nobody talks about it. Great point.