Pick Vidu Q4 Preview for native audio by default, five resolution tiers and clips up to 16 seconds. Pick Kling 3.0 to lock both the first and last frame.
Vidu Q4 Preview is live on HiAPI. Open the image-to-video model page, upload a first frame and generate; new accounts get 200 credits.
Vidu Q4 Preview vs Kling 3.0 at a glance
| Vidu Q4 Preview | Kling 3.0 image-to-video | |
|---|---|---|
| Model ID | vidu-q4-preview/image-to-video | kling-3.0/image-to-video |
| Image input | One first frame in image_url | 1–2 images in image_urls; the second is the last frame |
| Duration | 3–16 seconds, default 5 | 3–15 seconds, default 5 |
| Resolution | 540p / 720p / 1080p / 2K / 4K, default 720p | 720p / 1080p / 4K, default 1080p |
| Native audio | audio defaults to true; turning it off keeps the price | sound defaults to false; enabling it raises the rate |
| 720p per second | $0.109 | $0.1 silent / $0.143 with audio |
| 1080p per second | $0.1375 | $0.129 silent / $0.193 with audio |
| 4K per second | $0.446 | $0.479 |
Vidu Q4 bills the whole seconds you submit times the per-second rate for that resolution. The audio switch does not change the price: 5 seconds at 720p is $0.545, 10 seconds $1.09, 16 seconds $1.744. Kling 3.0 also bills per second: $0.1 and $0.143 at 720p, $0.129 and $0.193 at 1080p. Its 4K transformation tier is $0.479 per second.
Same prompt, one generation each
Each pair shares one first frame, 5 seconds and 720p, one generation each. Audio is on for the first pair and off for the second.
Seaplane: when the push-in lands and how the propeller blurs
| Vidu Q4 Preview (with audio) | Kling 3.0 (with audio) |
|---|---|
 |  |
Same prompt, same settings, one generation each.
Vidu Q4 shows a clear push-in by 2.5 seconds, motion blur on the propeller, and rings of ripples around the floats. Kling 3.0 holds a composition almost identical to the first frame at 2.5 seconds; its push-in only becomes obvious at 4.8 seconds, the propeller blur arrives late, the pilot's arm stays raised, and the float and dock lines shift between frames. Both files are 1280×720 at 5.04 seconds, with an AAC track.
The pilot waves, the seaplane engine starts and the propeller spins up, gentle ripples spread across the misty lake as the camera slowly pushes in, golden morning light
Generate with this prompt →. After it opens, upload the seaplane first frame, choose 5 seconds and 720p, and leave audio on.
Barista: where the shot ends and whether steam shows
| Vidu Q4 Preview (silent) | Kling 3.0 (silent) |
|---|---|
 |  |
Same prompt, same settings, one generation each.
Vidu Q4 returns 724×1268, keeping the first frame's 4:7 ratio. The camera ends on a close-up, the cup sits at the lower edge, and steam is visible above it. Kling 3.0 returns 720×1280 and keeps a wider framing at the end, with the window, pendant lights, and counter still in frame; no steam is visible in the cup. Both run 5.04 seconds, and both are silent in this pair.
The barista sets the milk pitcher down on the counter, the latte art in the cup settles into a leaf pattern and steam rises from it; she lifts her eyes to the camera and smiles. Cafe murmur and the hiss of the espresso machine. Slow handheld push-in, warm morning light through the window.
Generate with this prompt →. After it opens, upload the barista first frame, choose 5 seconds and 720p, and turn audio off so only the picture is compared.
Which one for which job
- Default audio without an extra switch → Vidu Q4 Preview
- A 540p draft pass → Vidu Q4 Preview
- One continuous 16-second shot → Vidu Q4 Preview
- 2K or 4K delivery → Vidu Q4 Preview
- 1080p with audio at $0.1375 per second → Vidu Q4 Preview
- A defined last frame → Kling 3.0
- 1080p silent at $0.129 per second → Kling 3.0
- 720p silent at $0.1 per second → Kling 3.0
What changes when you switch
Both models use POST https://api.hiapi.ai/v1/tasks with the same API key and return a taskId that you collect by polling or through callback.url. Four things change:
model: swapvidu-q4-preview/image-to-videoforkling-3.0/image-to-video.- Image field: Vidu Q4 takes one first frame as the string
image_url; Kling 3.0 takes an array inimage_urls, where one image is the first frame and two images add a last frame. - Audio field: Vidu Q4 uses
audio, on by default. Kling 3.0 usessound, off by default and billed at the audio rate when enabled. - Duration and resolution: Vidu Q4 covers 3–16 seconds across five tiers; Kling 3.0 covers 3–15 seconds across three.
Framing follows the input image on both, and neither model exposes aspect_ratio for image-to-video. Auth, callback URL, and polling stay the same.
FAQ
Can Vidu Q4 image-to-video take two images? No. It takes one first frame. For multiple references, use the Vidu Q4 reference-to-video model, with up to 15 reference images and 3 audio clips. For a last frame, use Kling 3.0.
Is audio on by default?
On Vidu Q4, audio defaults to true; turning it off returns a silent file at the same price. On Kling 3.0, sound defaults to false, and enabling it moves the request to the audio rate.
Does either reach 16 seconds? Vidu Q4 does, from 3 to 16 seconds. Kling 3.0 stops at 15.
How is framing set?
Both follow the input image. Vidu Q4 has no aspect_ratio field and returns 400 if you send one; Kling 3.0 exposes no such parameter.
Copy the seaplane prompt above, open the Vidu Q4 Preview model page, upload the first frame and run one 5-second 720p clip with audio. Then run the barista prompt with audio off.
Related: Vidu Q4 Preview image-to-video model page · Kling 3.0 image-to-video model page · Vidu Q4 Preview reference-to-video model page · image-to-video API docs
