Guide to Video Generation Basics
Learn how to create videos in getimg.ai: from writing prompts and adding references to adjusting basic settings like aspect ratio and length.
Step 1. Write a prompt
Log into getimg.ai, click "Video" and open "Create video" from Actions in the sidebar.
Then, describe what you want to see: be as brief or as detailed and technical as you’d like. Specify any dialogue, music, and ambient sounds you’d like to hear.
Examples:
- a surfer riding a wave at sunrise, cinematic lighting, sound of crashing waves
- a woman walking through a dimly lit art gallery, camera follows from behind, soft piano music playing, gentle reflections on polished floors.
getimg.ai’s smart prompt enhancement fills in the visual detail automatically, so even a short prompt will create a polished result.
Need more inspiration? Check out the Video Prompt Book.
Step 2. (Optional) Add References or Frames
You can guide your video using:
- First and Last Frame images: to define the exact start and end.
- Reference images: to show the AI what should appear later in the video, like a person or object entering mid-scene.
- Reference videos and audio: on five models (audio on four of them), to guide the clip with a video file or a sound file instead of a still.
- Elements: to reuse a person, product, style, or other concept you saved once, by typing @ElementName. They work on models that accept reference images. See the Guide to Elements.
Add them in any of these ways:
- Click the "First Frame" or "Reference" icon in the prompt box and select an existing image (or upload a new one). After you add a First Frame, you'll be able to add a Last Frame as well.
- Drag and drop an image from the gallery into the prompt box
- Hover an image, click ⋯, open Use in video, then pick Use as first frame, Use as last frame, or Use as reference. For a video, pick Use as reference in video instead.
You can attach several references at once, or use one image for the First Frame and another for the Last Frame. It's possible to only set the First frame, but not only the Last frame. How many you can attach depends on the model you generate with, and on Auto it follows whichever model Auto picks.
On Seedance 2.5, MiniMax H3, MiniMax H3 Max, and Wan 3.0, a reference can also be a video or audio file. You can attach up to 3 videos and up to 3 audio files per generation, alongside your reference images. Each attached file can run up to 15 seconds, audio as well as video, except on Seedance 2.5, which takes 30 seconds of either. Kling 4.0 takes video references too: up to 5 clips, 30 seconds in total, alongside up to 10 reference images, but no audio files.
Set the image purpose (Reference/First Frame/Last Frame) by selecting it from the dropdown under the thumbnail in the prompt box.
Learn more in our Video References Guide and First & Last Frame Guide.
Step 3. Choose basic settings
Below the prompt box, set the:
- Aspect ratio
- Length
- Variations: click the + and − icons to set how many clips to generate, from 1 up to 4.
On Auto, the best model is selected for you automatically.
To pick a specific model or turn off prompt enhancement, click the Auto / model name button in the bottom-left of the prompt box.
See our Video Generation Parameters Guide for more details.
Step 4. Generate
Click the arrow next to the prompt box to start the generation. While your clip is being created, you can immediately queue another.
Once your videos appear in the gallery:
- Hover over a result to like, delete, download, add to folder, or reuse the prompt
- Click a video to open the full view: from there, you can copy its prompt, see generation details (author, aspect ratio, length, model), and edit the clip.
Try experimenting with different prompts or reference setups to see how each change affects the motion and style.
Edit a Video from Full View
Want to change a clip you already have? Click any video in your gallery or a folder, including videos you uploaded, to open it in full view. The prompt box below already has that video attached as a reference, so you can describe the change right away (for example, “make it night” or “change the jacket to red”) and click the arrow.
In the model picker, you can choose from the available models that support videos as references. The edit comes back as a new video, and your original stays untouched.
Frequently Asked Questions
No. getimg.ai uses smart prompt enhancement, which automatically fills in visual details for you. A simple description of the subject and action (e.g., “a surfer riding a wave at sunrise”) is usually enough to create a high-quality video.
Make sure the model is set to Auto or a model that supports native sound (look for a speaker icon next to its name on the model list). Then, describe the audio directly in your text prompt. You can specify dialogue, music genre, or sound effects (e.g., "soft piano music playing, sound of crashing waves").
No, you must choose one method per generation. Use First/Last Frames if you need to lock the exact starting or ending image of the clip. Use Reference Images if you want to guide the AI on what (characters, objects) should appear inside the video content.
Yes, on Seedance 2.5, MiniMax H3, MiniMax H3 Max, Wan 3.0 and Kling 4.0. On the first four you can attach up to 3 video references and up to 3 audio references per generation, and use them alongside reference images in the same video. Each attached file can run up to 15 seconds, audio as well as video, except on Seedance 2.5, which takes 30 seconds of either. Kling 4.0 takes up to 5 video references, 30 seconds in total, alongside reference images but no audio. Every other video model takes reference images only.
Auto is recommended for most users because it automatically selects the best model for your prompt and keeps prompt enhancement on. Click the Auto / model name button in the bottom-left of the prompt box to pick a specific video model yourself or turn off prompt enhancement for full control.
Yes. The Variations control below the prompt box starts at 1, and the + and − icons step it up or down to any number up to 4, so you can ask for 3 clips just as easily as 2 or 4. Each run generates that many versions of your prompt in a single batch, so you can pick the best result. You can also queue new generations while others are still processing.