Image-to-Video Tutorial: Turn One Image into an AI Video
Image-to-video tutorial: turn a still image into AI video using first/last frames and reference images. Steps, models, costs, 7-day retention, TWD pricing.
Turning a still image into a moving video is called image-to-video. This tutorial shows you how to use first and last frames and reference images to generate a short video with AI.
What Is Image-to-Video
Unlike "text-to-video," where you give a sentence description and get a video, image-to-video takes an image as input. The AI generates motion based on your picture, which gives you more control over composition and subject.
First/Last Frames vs. Reference Images
- First and last frames (frame images): These set the first and last frames of the video, and the AI fills in the motion in between. This suits "change from this picture to that one."
- Reference images: These guide the style and texture and are not tied to the start or end. You can add several.
In Video Studio, these are two separate upload areas. You can use them independently, and only the fields you fill in are sent.
Step-by-Step Guide
- Go to Video Studio and switch to "Image → Video."
- Upload the first and/or last frame under "Start image." If you want style guidance, upload one or more images under "Reference images."
- Fill in the motion description (optional), then choose the model, duration, resolution and aspect ratio.
- Check the estimated cost, click "Generate video," and download the video when it's done.
How to Choose a Model
Not every model accepts image input. When choosing, check the specs on each model page. If you want fast output and good value, choose Seedance; if you care most about visual quality, choose Veo. The TWD price per second for each model is listed on its model page.
Cost and Retention
- Pricing is per second. The estimate is shown before generation, and the actual charge is shown when it completes.
- Generated videos are kept in the cloud for 7 days, so download them promptly.
- Pricing is in TWD, and uniform invoices are available.
Want to compare models first? See the tools comparison. When you're ready, start generating.
FAQ
How many seconds of video can one image generate?
It depends on the model. Common lengths are 5 to 10 seconds. Longer videos cost more, and the platform shows an estimate before generation.
What is the difference between first/last frames and reference images?
The first and last frames set the starting and ending pictures of the video. Reference images guide the overall style. The two can be uploaded separately and are independent of each other.
Which image formats are supported?
JPG and PNG are commonly supported, with a per-file size limit (about 10MB). After upload, the image can be used directly as generation input.
TWD billing · Taiwan invoices · leading AI models · OpenAI-compatible API