Luma Modify Video: adhere, flex and reimagine modes vs a Sume prompt

Luma's Modify Video endpoint has a required mode with nine values in three families: adhere 1 to 3, flex 1 to 3 and reimagine 1 to 3. They set how closely the output follows the source. Sume's video edit has no mode field: the request is a video_url plus a prompt, and how much changes is decided by what the prompt says.

What the three families mean

Luma describes adhere modes as keeping the output very close to the source, for subtle changes and light stylization. Flex modes adhere somewhat to shapes, characters and details and allow larger style shifts while keeping recognizable elements. Reimagine modes adhere much more loosely, for fundamental transformations and new aesthetics.

How to get the same range on Sume

The Sume edit example in the Video Router docs shows the pattern: Replace the bottle with an apple. Keep everything else the same. The first sentence is the change, and the second sentence is the strictness. There is no numbered dial, so write the constraint into the prompt.

A translation habit

A rough translation, as a starting point and not a rule that Sume documents:

• For an adhere-style edit, name one change and say to keep everything else the same.
• For a flex-style edit, name the style or look and list what must stay, such as the character and the framing.
• For a reimagine-style edit, describe the new look and say what you do not care about keeping.
• Run a short test and compare before you spend on a long source.

Keep each edit small

Sume's edit mode takes resolution as an optional field and refuses aspect_ratio and duration, so the output length follows the source. If you need a precise change, make one edit per request and chain the results, rather than packing several changes into one prompt.