> For the complete documentation index, see [llms.txt](https://animotive.gitbook.io/animotive-kimodo/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://animotive.gitbook.io/animotive-kimodo/getting-started/prompting-with-animotive-kimodo.md).

# Prompting with Animotive Kimodo

Once your Level is prepared for Animotive Kimodo, you're ready to start prompting. A **text prompt** is all you need for a first generation; **constraints** give you finer control when you want to pin a pose or position at specific moments.

## Text prompts

1. On the **Animotive Kimodo** track, click the **Add Section** button to add a new prompting section on the track.

   ![Add Section button on the Animotive Kimodo track](/files/5jKkaLofMRtopmLj5X1r)
2. Move the section and adjust its length like any other clip — its span sets the timing of the generated motion.

   ![Adjusting the prompt section length on the track](/files/wP5BnbdAdG4RXiRH2zJX)
3. Double-click the section to open the **Animotive Kimodo Prompt** window.
4. Enter your text prompt and click **Submit**.

   ![Animotive Kimodo Prompt window with a text prompt entered](/files/rDuaoqlLTHIZ7PoOHugd)
5. Click **Generate with Kimodo**.

   ![Generate with Kimodo button in the Sequencer toolbar](/files/zxjntx61XlfDn9uhVqs1)

The request appears in the **Animotive Kimodo Monitor** (**Tools → Animotive Kimodo → Monitor**) and, once complete, the animation is imported and placed on your character automatically.

## Kimodo Constraints

{% embed url="<https://youtu.be/Jg-V07d_XV8>" %}

Kimodo constraints are essentially keyframes that help enforce the character's pose or position at key moments during the generated animation. They live on the companion **Kimodo Constraints** track and come in six channels:

| Constraint                    | What it does                                                                                                                    |
| ----------------------------- | ------------------------------------------------------------------------------------------------------------------------------- |
| **Full-Body**                 | Enforces the character's full body pose at that moment. Can also be used to guide or influence the overall generated animation. |
| **2D Root**                   | Influences the character's root position at key moments. Note: it does **not** consider the Z (up/down) axis.                   |
| **Left Hand**, **Right Hand** | Influence the position of these hand end effectors.                                                                             |
| **Left Foot**, **Right Foot** | Influence the position of these foot end effectors.                                                                             |

### Keyframe both tracks

When working with Kimodo Constraints it is **crucial to keyframe both** the desired **Kimodo Constraints** track **and** the **Control Rig** track. The Kimodo Constraint is what gets submitted to the Kimodo request in the backend, but it **samples its pose from the Control Rig track**. A constraint keyframe with no matching Control Rig pose has nothing to read.

![A constraint keyframe alongside its matching Control Rig keyframe](/files/nJnlqQ4ST3U3MPQbyyoW)

### Holding a pose

If you want the character to **hold** a pose or position, add duplicates of that constraint at several moments across the section, rather than relying on a single keyframe.

![Duplicated constraint keyframes across a section to hold a pose](/files/arzWizQlQKioWmJsxnQR)

For the full channel reference and authoring patterns, see [Constraints Track](/animotive-kimodo/working-in-sequencer/constraints-track.md).

## Prompting best practice

A few things make a big difference to your results:

* **Start the prompt with a subject.** Write *"a person runs forward"* rather than *"running forward"*. Kimodo responds best to prompts phrased as a full action with a subject.
* **Aim for 2–10 seconds.** That's the sweet spot for section length. Very short or very long sections produce weaker results.
* **Keep it focused on the action.** The model doesn't need scene-setting or camera direction — just what the character does.

NVIDIA's Kimodo team publishes excellent prompting guidance, including phrasing and known limitations. It's worth a read: [NVIDIA Kimodo — Limitations & Key Concepts](https://research.nvidia.com/labs/sil/projects/kimodo/docs/key_concepts/limitations.html)

## Next

* [The Animotive Kimodo Track](/animotive-kimodo/working-in-sequencer/animotive-kimodo-track.md) — multi-section authoring, transition frames, and seed control.
* [Constraints Track](/animotive-kimodo/working-in-sequencer/constraints-track.md) — the full six-channel reference.
* [Import & Results](/animotive-kimodo/working-in-sequencer/import-and-results.md) — what happens to the generated animation once it returns.