Luma Dream Machine beside Step-Video-TI2V, point by point
The docs of Luma Dream Machine and Step-Video-TI2V part on 3 of the six camera points: camera moves, frames and motion amount. They read alike on framing, cuts and negative prompt. As of 2026-09-25.
| Field | Luma Dream Machine | Step-Video-TI2V |
|---|---|---|
| How a camera move is asked for | Commands or presets | Words in the prompt |
| Shot sizes and angles in the docs | None in the docs | None in the docs |
| Cuts inside one generation | None in the docs | None in the docs |
| Start and end frames | First and last frame | First frame only |
| A negative prompt | None in the docs | None in the docs |
| A setting for how much the picture moves | None in the docs | A motion setting |
Inclusion rule. The six points every model on the register is read on. Order. The order the points take on every model page.
1How the two differ
Luma Dream Machine reads “Commands or presets” on camera moves, while Step-Video-TI2V reads “Words in the prompt” on camera moves. Luma Dream Machine reads “First and last frame” on start and end frames, while Step-Video-TI2V reads “First frame only” on start and end frames. Only Step-Video-TI2V documents a motion setting. Luma Dream Machine's docs use 6 of the site's camera words and Step-Video-TI2V's use 7, sharing pan, zoom and orbit or arc.
2The differing points, in the docs' words
How a camera move is asked for
Camera is controlled by language in Dream Machine.
Luma AI, Dream Machine API video generation
Supported Camera Movements
StepFun, Step-Video-TI2V repository
Start and end frames
With start and end keyframes
Luma AI, Dream Machine API video generation
To incorporate the image condition as the first frame of the generated video, we encode it into latent representations
StepFun, Step-Video-TI2V repository
A setting for how much the picture moves
Luma Dream Machine: the docs read do not cover this.
The default motion_score = 5 is suitable for general use.
StepFun, Step-Video-TI2V repository
- Camera motionsCamera moves are words in the prompt, taken from a list the API returns
- Camera controlA table of supported camera movements, from a fixed camera to orbit shots and rack focus, asked for in the prompt
- KeyframesKeyframes set a start image, an end image or both
- Model summaryThe image condition becomes the first frame of the generated video
- Motion dynamics controlA motion_score setting, 5 by default; 2 for more stability, 10 or 20 for more intense action
3Luma Dream Machine, Step-Video-TI2V and a third set of camera docs
| Camera point | SceneMixer | Source |
|---|---|---|
| How a camera move is asked for | Camera moves are written into each shot's text: fixed, handheld, tracking or a slow push in | SceneMixer, how to write one shot |
Inclusion rule. Points where the two models above differ and SceneMixer's guides say something; points its guides leave open are not shown. Order. The order the points take on every model page.