How to Make AI Videos: Prompts, Costs and a Real Example
Create a short AI video from text or a photo, check the starter-credit limits, review a real published example and assemble clips in your editor.
To make an AI video, choose one shot, write a clear motion prompt, select compatible model settings and review the actual output before downloading it. You can begin with text or an existing photo. Multiple shots, narration and captions usually need a separate editing step.
This guide explains the workflow and includes an existing published example. Suggested prompts below are starting points, not claims about results we have measured.
Start with one visible action
Describe the subject, its action, the setting and a camera movement. For a product photo, you might try:
A slow camera orbit around the ceramic mug. Keep the handle, proportions and surface pattern consistent. Soft side light, a steady background and one continuous shot. No added text or logo.
Use image-to-video when a source picture matters. Use text-to-video for an invented scene. Neither instruction guarantees that every detail will remain unchanged, so inspect the result against the source.
Choose settings and review the quote
Check the selected model’s available duration, resolution and aspect ratio. A model family’s official documentation can describe features that are not exposed by a particular hosted route. The model directory and generator show the settings configured on this site.
Generating requires an account and credits. At the prices checked on October 18, 2026, new accounts receive 30 welcome credits; one 5-second, 480p Seedance 1.0 Pro Fast clip costs 30 credits. Other models use purchased credits. Use current pricing for package and expiry rules, and review the exact quote before confirming.
Model names alone do not establish availability. In particular, the Sora 2 page explains the distinction between a third-party supplier option and OpenAI’s official API retirement notice.
A real published video example
The following sample comes from the paid Vidu Q4 reference workflow. Its source was a synthetic adult portrait. The generated output and downloaded application file were checked against the same stored file.
| Recorded item | Value |
|---|---|
| Generated | October 18, 2026 |
| Model and input | Vidu Q4 reference mode, one synthetic portrait |
| Requested settings | 3 seconds, 720p, 9:16, native audio |
| Delivered file | About 3.04 seconds, 720 × 1280, audio track present |
| Site charge for this recorded run | 180 credits |
The review sampled frames and checked that the audio track was non-silent. It did not assess speech quality or establish a success rate. Results depend on the source image and settings. Current prices may differ from the recorded run.
Generate, review and change one thing
- Open the tool, prepare a prompt and add compatible input files.
- Check the settings and credit quote, sign in and confirm the generation.
- Wait for the completed result. Do not start a duplicate task just because a request acknowledgement is delayed.
- Watch the clip. Check subject consistency, unwanted objects, motion and audio.
- If you retry, change one part of the prompt or one setting so you can understand the difference.
There is no fixed number of attempts that guarantees a usable clip. If a task fails, the site’s purchase rules describe credit compensation; credit compensation should not be confused with a cash refund.
Assemble clips into a finished video
Download the completed MP4. In your editor, arrange the shots, add narration and captions, set pacing and check the final export. For social posts, verify the destination’s current upload requirements and preview the crop.
If you begin with written material, Article to Video Clips can suggest individual shots. AI B-Roll creates supporting footage, and the visual hook tool focuses on an opening shot. These tools do not automatically assemble a finished narrated video.
FAQ
Can I make an AI video for free?
The current welcome allowance covers one eligible starter clip as described above. You need an account to generate; other models use purchased credits. Check the live quote before submission.
Should I upload a photo or use text?
Use a photo when the existing subject or composition matters. Use text when you want to invent a scene. Review the output in either case.
Is the example a benchmark for every model?
No. It is one published result from a specified workflow and recorded settings. It does not predict the quality of another prompt or model.
Can the generator produce my complete article video?
It creates individual clips and shot suggestions. Use an editor for narration, captions, music and assembly.
Start with a small, clearly defined shot. Verify the settings, cost and output, then build a longer video from clips you have actually reviewed.