Vidu Q4 Turns Your Images Into 4K Video With Sound
Last verified: November 3, 2026
Vidu Q4 is Shengshu Technology's next-generation flagship video model. Its standout mode is Reference-to-Video, which combines up to 15 reference images and 3 voice references into one clip and keeps characters, styles, scenes and voices aligned. Shengshu released it to the public on October 7, 2026 as Q4 Preview. This first public preview came out ahead of the full Q4 model and is built for expressive audio and video. The model also has an Image-to-Video mode, which animates a single image from a text prompt.
Voice is central to how it works. Each voice reference clones a character's voice with natural emotion and lip-synced delivery. Q4 Preview also coordinates facial expression, emotion, body movement and voice, which gives more nuanced expressions and more natural movement. It improves camera work and visual effects too. The model goes up to 4K resolution and 16-second clips. Q4 Preview's 2K and 4K output has 10-bit color depth.
Q4 Preview is an early release. Shengshu says feedback from creators on performance, creative control and everyday production needs will help shape the full version. You can run both modes on this page.
Explore Vidu AI's Models
Output, Length and Reference Limits
What each run accepts and returns in Image to video and Reference to video.
Output resolution
540P, 720P, 1080P, 2K or 4K (720P by default)
Clip length
3 to 16 seconds, 5 by default
Reference images
1 to 15 in Reference to video; 1 in Image to video
Voice references (Reference to video)
Up to 3 MP3 clips, each up to 12 s and 25 MB
Reference to video frame shapes
16:9, 9:16, 1:1, 4:3 or 3:4 (16:9 by default)
Image to video frame shape
Same aspect ratio as the uploaded image
Check These Before You Hit Generate
Six checks that prevent mismatched framing, muddled references and rushed performances.
Crop the image to the shape you'll publish
In Image to video, the clip uses your picture's proportions, so crop to vertical, widescreen or square before you upload. Use a JPG, JPEG, PNG or WEBP file that is at least 300×300 pixels and no larger than 10 MB.
Give each reference image one clear subject
In Reference to video, put the character, outfit, product and location in separate, well-lit images instead of one busy collage. Then describe each of them in the Prompt so your text and your references refer to the same things.
Prepare clean voice clips of 12 seconds or less
Add Audio accepts up to three MP3 files, each up to 12 seconds and 25 MB. Use one speaker per clip with no music underneath, because each clip is there to carry one character's voice.
Direct the performance, not just the movement
Write the emotional shift, a gesture or two, and each spoken line in quotes along with who says it. A prompt that only describes movement wastes the model's strength in matching expression, body language and voice.
Allow enough seconds for dialogue and action
Video Duration starts at 5 seconds. Increase it, up to 16, when a scene has a spoken line and a reaction or more than one shot. Otherwise the delivery can feel rushed or a camera move can get cut off.
Test at 540P or 720P before a 4K run
Run at a low resolution first to check that the references, the performance and the camera come across as intended. Every run is a new take, so your 2K or 4K run will be a fresh generation from the approved prompt, not an upscale of the draft.
Vidu Q4 or Kling O3: Match the Generator to Your Shot
Each row shows what the Vidu Q4 and Kling O3 generators on Vidofy offer: their modes, uploads and settings. Use it to see which one suits your source material and delivery format before you generate.
| Feature/Spec | Vidu Q4 | Kling O3 |
|---|---|---|
| Modes you can run | Image to video, Reference to video | Reference to video, First to last frame, Image to video, Text to video |
| Video Duration options | 3–16 s, default 5 s (both modes) | 3–15 s, default 3 s (all four modes) |
| Resolution options | 540P, 720P, 1080P, 2K, 4K; default 720P (both modes) | 720P, 1080P, 4K; default 720P (all four modes) |
| Image uploads in Reference to video | Add Image: 1–15 images | Image: 1 required, plus Add elements: 1–7 images |
| Audio reference upload (Vidu Q4 Reference to video) | Add Audio: up to 3 MP3 clips, each up to 12 s and 25 MB (optional) | The Kling O3 Reference to video generator shows no audio upload |
| Aspect Ratio in Reference to video | 16:9, 9:16, 1:1, 4:3, 3:4; default 16:9 | 1:1, 9:16, 16:9; default 1:1 |
| Prompt length | Up to 5,000 characters | Up to 2,048 characters |
| Generate Audio switch | In both modes, on by default | In all four modes, on by default |
Where Each Generator Fits Your Production
Building from references or starting from text and end frames
Vidu Q4's Reference to video generator is the better fit for large casts and sets. It holds nearly twice as many images as Kling O3's reference setup, and it is the only one of the two reference generators here that accepts voice clips, so a character can come with both a face and a voice. Its larger prompt box also leaves room for several lines of dialogue and shot-by-shot direction.
Kling O3 offers starting points that Vidu Q4 does not have on this page: a text-only start and a mode that fixes both the opening and closing frame of a shot. If you have no source image, or the clip has to end on a specific frame, open Kling O3.
Finishing resolution, length and frame shape
Both generators go up to 4K and add sound by default, so final sharpness rarely decides it. Vidu Q4 adds a 540P tier for low-resolution test runs and a 2K step between 1080P and 4K. Its clips can also run one second longer.
For different platforms, Vidu Q4's reference mode adds 4:3 and 3:4 to the square, vertical and widescreen shapes both generators offer. That helps with feed posts and portrait displays that crop awkwardly from 16:9 or 9:16.
Choose by Source Material and Delivery Format
Use this quick guidance to pick the best option for your workflow.
How to choose: Pick Vidu Q4 Reference to video when a scene needs more than eight reference images, voice clips that carry a character's voice, a 4:3 or 3:4 frame, or a prompt longer than 2,048 characters. Pick Vidu Q4 in either mode when you want a 540P or 2K render or a 16-second clip. Pick Kling O3 when you're starting from text alone or need to set both the first and last frame. Both reach 4K, and both add sound with a Generate Audio switch that is on by default.
From Upload to Finished Clip
Five steps take you from choosing a mode to a finished video with sound.
Choose Image to video or Reference to video
In the model list, pick Vidu Q4 Image to video to animate one picture, or Vidu Q4 Reference to video to build a scene from several pictures and optional voice clips.
Upload with Image, or Add Image and Add Audio
For Image to video, put one picture in Image. For Reference to video, add your character, product and location pictures with Add Image, and attach MP3 voice clips with Add Audio if someone speaks in the scene.
Write the Prompt and try Enhance Prompt
Describe the action, the performance, any spoken lines and the camera in the Prompt box. Enhance Prompt, under the box, suggests fuller versions of a rough idea. Your text only changes if you pick a suggestion.
Set Video Duration, Resolution, Aspect Ratio and Generate Audio
Choose a length and a resolution. In Reference to video, also choose an Aspect Ratio. Leave Generate Audio on if you want the clip to have sound.
Click Generate
Each run returns one video. To get another take, change the prompt, references or settings and click Generate again.