Vidu Q4 Preview
Vidu Q4: animate one frame or build a scene from references
Vidu Q4 is the newest video model from ShengShu Technology, released as a public preview on October 7, 2026. On Kuca you can use it in two ways: bring a single image to life, or hand it a board of reference pictures and voice clips and let it stage a new scene around the same characters. Clips run from 3 to 16 seconds and render at anything from 540p up to 4K.
Vidu Q4 on Kuca at a glance
- Modes
- Image to video · Reference to video
- Clip length
- 3–16 seconds
- Resolution
- 540p · 720p · 1080p · 2K · 4K
- References
- Up to 12 images and 3 voice clips
- Aspect ratio (Omni Reference)
- 16:9 · 9:16 · 4:3 · 3:4 · 1:1
What Vidu Q4 changes
Most of what is new in Vidu Q4 is about acting. ShengShu says the model ties facial expression, body movement and voice together more closely than earlier Vidu releases, so a glance, a pause or a shout reads as one performance instead of a face pasted onto motion. It also pays closer attention to how the camera moves and cuts around that action, and busy effects such as sparks, smoke, fireworks and explosions sit in the scene with more weight and believable light.
Output goes up to 4K with 10-bit color, which helps with smooth gradients in skies, skin tones and night scenes. The release carries the Preview label: it is the first public version of a new flagship line, and for now it works only from images and reference material. There is no text-only mode yet, so every Vidu Q4 video you make on Kuca starts from at least one picture.
Two ways to start a video
One image
Image to video: one frame, one shot
Image to video treats your picture as the opening frame. Vidu Q4 keeps its composition, lighting and subject, then moves forward from there according to your prompt. Use it when you already have the exact frame you want: a key visual from a storyboard, a product photo, a portrait, or the last frame of an earlier clip that you want to continue. In the generator above, this is the Frame generation tab.
The clip takes its framing from your image, so crop the picture to the shape you need before uploading it. In the prompt, describe what happens next, such as the action, the camera move and the mood, rather than describing the picture again. The model can already see it.
Reference board
Reference to video: a board of references, a new scene
Reference to video does not lock the opening frame. You give Vidu Q4 the pieces of a scene, including characters, wardrobe, props, products and a location, and it composes new shots that use them. This is the mode for short dramas, recurring mascots and product stories, where the same person or object has to look right across several separate clips. In the generator, it sits under the Omni Reference tab.
Here you pick the aspect ratio yourself, and you can switch on audio so the model adds dialogue and sound effects to the picture. Optional voice clips tell it how each character should sound. Everything you upload is a reference rather than a frame, so the model is free to choose angles, distance and blocking.
Building a reference board that holds up
Kuca accepts up to 12 reference images for each Vidu Q4 generation; Vidu lists up to 15 on its own platform. More pictures do not automatically mean a better result. Each image should earn its place by showing something the model cannot guess: a face from the front and from a three-quarter angle, the complete outfit, a prop held the right way, the room or street where the scene takes place.
Use clean, evenly lit pictures with one clear subject each. A crowded group photo forces the model to work out who matters. If a character has to look the same in a later clip, reuse the same images in the same order and describe them plainly in the prompt, for example the woman in the red coat or the silver thermos, so that the text and the pictures point at the same things.
- Face, front
- Face, ¾ view
- Full outfit
- Prop
- Location
- Second character
Up to 12 images
Voice clips
- Lead
- Partner
- Narrator
Voice clips and sound
Reference mode takes up to three voice clips in MP3 format, each between 3 and 12 seconds long. They keep a character's voice and delivery steady from one shot to the next, which matters when a series is edited together from separate generations. A clip of clear speech with no music underneath works best, and one speaker per file keeps the voices from blending.
Voice clips only shape how speech sounds. To get speech at all, switch on Generate audio and write the line in your prompt, together with who says it and how: whispered, out of breath, half laughing. With audio on, Vidu Q4 also adds ambience and effects that match the picture. Leave it off if you plan to score, dub or mix the clip yourself later.
Length, resolution and credits
Every Vidu Q4 clip can be anywhere from 3 to 16 seconds long. Short clips are easier to direct: one action, one camera move, one emotional beat. Save the longer end for scenes that need a setup and a payoff, or for prompts that ask for a cut inside the clip, and remember that a longer clip is also a longer wait.
Resolution has the biggest effect on cost. 540p is enough to check timing, framing and whether your references are being read correctly. Move up to 720p or 1080p for social posts, and keep 2K and 4K for the version you will actually deliver or color grade. The Generate video button shows the exact credit cost for your length and resolution before you spend anything, and switching audio on does not change it.
Prompts that direct a performance
Write the prompt like a short shot description: who is in frame, what they do, how they feel and what the camera does. She reads the message, freezes, then slowly looks up at the door, with a slow push-in and warm lamp light, gives the model a beat to perform. A string of adjectives such as cinematic, emotional and beautiful does not give it anything to act on.
Keep one idea per clip. If you need a reaction shot and a wide shot, generate them separately with the same references and cut them together afterwards. When something goes wrong, change one thing at a time, whether that is the wording of the action, a single reference image or the length, so you can tell which change fixed it.
The detective from the first two images sits across the table from the woman in the red coat. He slides a photo toward her. She glances at it, looks away and says quietly, "I've never seen him before." Slow push-in on her face, rain running down the window behind them.Vidu Q4 questions
How do I make a Vidu Q4 video on Kuca?
Open the generator at the top of this page and choose a mode. Under Frame generation, upload one picture and describe the motion. Under Omni Reference, add 1 to 12 reference images, up to 3 MP3 voice clips if you want them, and describe the scene. Pick the length and resolution, check the credit cost on the button, then generate. The finished clip appears in your history.
Should I use image to video or reference to video?
Use image to video when you already have the exact first frame and simply want it to move. Use reference to video when the same character, outfit or product has to appear in new shots that you describe, or when you want dialogue in a consistent voice.
Can Vidu Q4 make a video from a text prompt alone?
Not in the current preview. Both modes need at least one image. If you only have an idea, create a key frame first with an image model, then animate it with Vidu Q4 or use it as a reference.
How do I keep the same character across several clips?
Generate each clip in reference mode with the same set of images in the same order, describe the character with the same words every time and reuse the same voice clip. Keep lighting and wardrobe consistent in the reference images themselves, because the model follows what it sees.
What kind of voice clip works best?
An MP3 between 3 and 12 seconds long with one person speaking clearly, no music and little room echo. Record it in the tone you want, since a calm clip produces a calm voice. Then switch on Generate audio and put the actual line in quotation marks in your prompt.
How long can a clip be, and what does it cost?
From 3 to 16 seconds, at 540p, 720p, 1080p, 2K or 4K. The cost is calculated per second and rises with resolution, and the Generate video button shows the exact credits for your settings before you start. Turning on audio does not add to the price.
Related tools
Try Vidu Q4Image to Video
Animate a photo with whichever model suits the shot, including faster and lower-cost options.
MiniMax H3 Max
Another reference-driven model that also accepts video clips as references and offers preset camera moves.
Product to Video
Turn product photos into short ads and showcase clips for stores and social feeds.