AI Video References: Choose Action, Camera Moves and Simple 3D Clips

Play the whole clip to inspect how the action starts, develops and ends. These AI-generated clay-style reference clips play when you press the player controls.

These clips illustrate what to inspect. Check your tool’s input requirements before using them: character motion controls may not support cars or two-person interactions.

A clay render uses simple models to show movement, layout and camera motion. AI can produce a similar look without an editable 3D project. Depth video represents distance from the camera; lighting and shadows on a gray surface are not depth. Name the versions separately and keep a record of their source.

After choosing a reference, prepare the input using your platform’s instructions. These links lead to each platform’s own guide.

• Kling: character motion control guide — Character images and motion reference videos.
• Higgsfield: using Kling motion control on its platform — Character images, motion references and framing; check your current version.
• LTX: depth, pose and edge control — A workflow for the corresponding open-source models, not a universal upload option.

A clip carries several kinds of information. Decide which one matters for this shot.

Use this table to choose a clip, then check what your tool supports. A character image may show what someone looks like, while a motion video shows how they move.

Video-to-video broadly describes generating or changing imagery from an existing video. Motion reference workflows focus on borrowing performance or movement, often alongside a separate character input. Tools may provide separate entry points for these tasks.

Runway Act-Two, for example, uses a performance video alongside a character image or video. What you can control depends on which type of character file you choose. Runway documentation

Before trimming a clip, check where the tool asks you to upload it.

A simple 3D scene lets you plan where objects go and how the camera moves before adding detail. Blocks can stand in for a table and product while a camera passes alongside them. Export the scene as video to inspect when the product enters the frame and where it ends up.

Seedance 2.5 describes using simple 3D videos without surface detail to guide object placement, poses, movement and camera angles. Complex movements may still differ from the reference. Seedance introduction

Whether you build the scene yourself or ask an AI tool to help, review the exported video for camera placement, movement and timing. Organize those videos in ShotMind and keep the 3D project files in your modeling software. These simple videos have fewer visual details, so try searching your own footage to see what works. You can also browse and play clips to choose one.

Suppose you want AI to generate a person reaching for a cup, with a natural hand movement. You can find a similar action in a video you have saved, trim that clip, and use it in a tool that accepts motion references.

1. Import your saved video into ShotMind. Use footage you own or may manage. ShotMind detects shots; you choose which ones to analyze. If the footage is already analyzed, go straight to search.
2. Search for the action you need. For example, type “a person reaches for a cup on a table” to look for possible matches in analyzed shots.
3. Play the clip and check the whole action. Look for the reach, grip and lift, and check whether the hand is hidden. A similar-looking frame is not enough; watch the movement from start to finish.
4. Trim the reference in your video editor. Open the original video in your usual editor, keep the action you need, and match the generation tool's duration and aspect-ratio requirements.
5. Upload it in your generation tool. Choose a feature that accepts motion references. Upload the trimmed video and add the character image, text or other material that tool requires, then generate. You find footage in ShotMind and generate the new video on the chosen platform.
6. Save the reference and result. Note which result used which reference in your own folders or spreadsheet, so you can find it again when you need a similar action.

Describe the image you want to find. These example phrases give you a starting point; change the person, action or setting to suit your search.

• Action: a person reaches for an object on a table, with a clear hand movement.
• Camera: the camera passes beside a subject, gradually revealing the space behind it.
• Mood: a cool industrial setting, strong side light, restrained movement.
• Composition: a small person in a wide environment with plenty of negative space.

A result may match only part of your description. Play it before choosing it, especially if you need a particular camera direction or a complete action. Learn how to organize and search video references

Full videos stay on your Mac. Only the shots you choose to analyze are sent to the cloud. The results and preview images are saved on your Mac for later browsing and search. Searching and browsing analyzed content use no credits, called Points. Analyzing new shots uses Points.

ShotMind helps manage and retrieve references. Generation, character settings and platform parameters remain in your generation tool.