ShengShu · Vidu Q4 Preview · New on Visiojoy, October 2026
Turn a photo or a set of reference images into a video with dialogue and sound, up to 4K. Upload your images below, describe the scene and generate it in your browser.
Vidu Q4 is the latest video model from ShengShu Technology, the team behind the Vidu AI video app. ShengShu released it as Vidu Q4 Preview on October 7, 2026, following Vidu Q3. Its focus is character performance: faces, gestures and voices that move together, with audio generated in the same pass as the picture.
Vidu Q4 always starts from images. Animate a single photo, or combine up to 15 reference images and 3 voice clips to keep characters, products and places consistent. Output goes from 540p to 4K in clips of 3 to 16 seconds. Visiojoy is an independent platform, not affiliated with ShengShu or Vidu.
Four things that set it apart from other video models on Visiojoy.
ShengShu built Q4 around character performance: expression, emotion, body movement and voice moving together. Write the small beats — a hesitation, a glance away, a laugh that fades — into the prompt. Useful for short films, skits and talking-head ads.
Add up to 15 images — characters, outfits, products, locations — and refer to them by number in the prompt. The model keeps each one recognizable while it invents new angles and actions.
Audio is generated with the picture: dialogue, effects and ambience. Add up to 3 short MP3 voice references so a character sounds the way you want, with lip movement matched to the line.
Five resolutions from 540p at 8 credits per second to 4K at 63. Test motion cheaply, then re-run the keeper at 4K — still well under Seedance 2.0's 208 credits per second at 4K.
Two ways to start a clip — both need an image. The generator above picks the mode from what you upload; the audio switch and full settings are in the AI Video workspace.
You upload
1 image
The video starts from your photo and keeps its aspect ratio. A prompt is optional — leave it empty to let the model decide, or describe what moves and what we hear.
You upload
1–15 images, up to 3 MP3 voice clips
References set who and what appears; the prompt sets the scene. Choose 16:9, 9:16, 1:1, 4:3, 3:4. Voice clips need at least one image next to them.
Q4 Preview is ShengShu's step up from Vidu Q3. The big jumps are resolution and reference control; the trade-off is that it no longer starts from text alone.
| Vidu Q3 | Vidu Q4 Preview | |
|---|---|---|
| Resolution | Up to 1080p | 540p to 4K, 10-bit color |
| Clip length | Up to 16 seconds | 3–16 seconds |
| References | Multi-image reference mode | Up to 15 images plus 3 voice clips |
| Audio | Native audio and dialogue | Native audio, voices guided by your clips |
| Focus | General scenes and storytelling | Acting, expressions and action shots |
| Start from text only | Yes | No — image or references required |
1
Use + to animate one photo, or switch to Reference to Video and add up to 15 images and 3 voice clips. Vidu Q4 always needs at least one image.
2
Say who does what, how they feel, how the camera moves and what we hear. Refer to references as image 1, image 2, audio 1. Put spoken lines in quotes.
3
Pick 540p to 4K and a length of 3–16 seconds. The credit cost shows before you start; download the finished video from the AI Video workspace.
Vidu Q4 prompts written for images and references. Click one to load it into the generator, add the images the card mentions, then match the settings shown.
Upload a portrait or a character sheet and describe how the person feels and reacts. Close and medium shots let the facial performance read.
The man in the photo answers his phone in a dim kitchen. His polite smile fades, he sets down his mug and slowly sits, staring at nothing. Static medium shot, then a slow push-in. Sound: a ticking clock, a muffled voice on the line.
The girl from the image stares at her laptop screen, eyes widening, then jumps out of her chair and screams with joy, arms in the air. Handheld camera following her. Sound: her scream, a chair rolling back.
Image 1 waits on a train platform. Image 2 steps off the train; they recognize each other, hesitate, then hug and laugh. Slow tracking shot along the platform, golden evening light. Sound: train brakes, station announcements, their laughter.
Use Reference to Video with one image per speaker. Add an MP3 voice clip (3–12 seconds) for each character whose voice should match.
Image 1, a detective, sits across from image 2 in a grey interview room. Image 1 slides a photo across the table and says in the voice of audio 1: "Tell me where you were on Friday." Image 2 stays silent, jaw tight. Alternating close-ups. Sound: fluorescent hum.
The chef from image 1 stands behind a kitchen counter full of fresh vegetables, smiles at the camera and says in the voice of audio 1: "Today we're making the fastest pasta you'll ever cook." Bright studio light, slight handheld movement.
Image 1, a grandmother, reads to image 2, a sleepy boy, in a warm bedroom lit by a bedside lamp. She whispers: "And the dragon finally fell asleep." He yawns and closes his eyes. Slow push-in. Sound: soft rain outside.
Upload clean product shots as references so the packaging and logo stay exact, then add a person or a setting image to build the scene around it.
The perfume bottle from the image stands on wet black stone as a slow-motion splash of water rises around it, light refracting through the glass. Camera slowly orbits. Sound: a deep whoosh, droplets falling.
The model from image 1 walks down a neon-lit Tokyo street wearing the jacket from image 2, turns to the camera and pulls up the collar. Tracking shot at waist height, light rain on the pavement. Sound: city ambience, an upbeat bass line.
Morning sun in the kitchen from image 2. A hand places the coffee bag from image 1 on the counter, scoops the beans into a grinder and steam rises from a fresh cup. Close-up macro shots. Sound: grinder, pouring coffee, birds outside.
ShengShu tuned Q4 for fast movement and effects like fire, smoke and explosions. Describe the move, the camera and the effect in that order, and keep each clip to one action beat.
The runner from the image sprints across a rooftop at dusk, vaults over an air vent and leaps to the next building. Camera tracks beside him, then whips to follow the jump. Sound: footsteps, heavy breathing, wind.
The mage from the image raises both hands; a spiral of fire grows between her palms and lights up the stone hall, embers drifting past the camera. Low-angle shot, slow push-in. Sound: a rising roar of flames.
Image 1 and image 2 duel with swords in a bamboo forest, blades clashing, leaves exploding around them. Wide shot, then quick close-ups of each strike. Sound: metal clanging, bamboo creaking in the wind.
All three are in the same model menu on Visiojoy. Here's when each one fits.
| Vidu Q4 | MiniMax H3 | Seedance 2.0 | |
|---|---|---|---|
| Top resolution | 4K | 2K | 4K |
| Clip length | 3–16s | 4–15s | 4–15s |
| Text only | No — needs an image | Yes | Yes |
| References | 15 images, 3 audio | 9 images, 3 videos, 3 audio | 9 images, 3 videos, 3 audio |
| Audio | Built in, can be switched off | Stereo, always on | Built in, can be switched off |
| Cost | 8–63 credits / s | 8–13 credits / s | 19–208 credits / s |
| Pick it for | Character scenes and 4K on a budget | Text-to-video in 2K with stereo sound | Video references and premium 4K |
Comparing more models? Browse all AI video models. Building an agent workflow? Vidu Q4 is also available through the Visiojoy MCP server (image-to-video).
You pay per second of video. Image-to-video and reference-to-video cost the same, audio on or off. Draft at 540p or 720p, then render the keeper at 1080p or above. The exact cost shows on the Generate button before you start.
| Resolution | Per second | 3s clip | 5s clip | 10s clip | 16s clip |
|---|---|---|---|---|---|
| 540p | 8 credits | 24 | 40 | 80 | 128 |
| 720p | 16 credits | 48 | 80 | 160 | 256 |
| 1080p | 20 credits | 60 | 100 | 200 | 320 |
| 2K | 31 credits | 93 | 155 | 310 | 496 |
| 4K | 63 credits | 189 | 315 | 630 | 1,008 |
Credit packs start at $9.90 for 640 credits — about 40 seconds of 720p or 10 seconds of 4K video. Reference images and voice clips are included in the price. See plans and credit packs.
Vidu Q4 is an AI video model from ShengShu Technology, the company behind Vidu. ShengShu released Vidu Q4 Preview on October 7, 2026 as the successor to Vidu Q3. It turns an image or up to 15 reference images into 3- to 16-second clips at up to 4K, with audio generated in the same pass. On Visiojoy it has been available since October 2026.
It is Vidu Q4 Preview, the version ShengShu has released so far. ShengShu has said the full Q4 will follow and may change specs and limits; this page and the generator update when it does.
Vidu Q3 tops out at 1080p. Vidu Q4 adds 2K and 4K output with 10-bit color, takes up to 15 reference images plus 3 voice clips to control how characters look and sound, and is tuned for facial acting and action shots. Both make clips of up to 16 seconds with native audio. Q4 Preview drops Q3's text-only and start-and-end-frame modes, so it always starts from an image.
No. Vidu Q4 offers image-to-video and reference-to-video, so every generation needs at least one image. Upload one photo to animate it, or add reference images in Reference to Video mode. For text-only prompts, pick another model such as MiniMax H3 or Seedance from the same menu.
540p, 720p, 1080p, 2K, 4K. Clips can be any whole number of seconds from 3 to 16. In Reference to Video you choose 16:9, 9:16, 1:1, 4:3, 3:4; in image-to-video the clip keeps your image's aspect ratio.
Up to 15 images (JPG, PNG or WebP, up to 50 MB each) and up to 3 voice clips (MP3 only, 3–12 seconds each, up to 50 MB). Vidu Q4 doesn't take reference videos. Audio can't be the only reference — add at least one image. Adding references requires signing in.
Yes. Audio is on by default: dialogue with matching lip movement, sound effects and ambience. You can switch it off in the AI Video workspace for a silent clip; the price is the same either way.
It's billed per second of output: 8 credits at 540p, 16 at 720p, 20 at 1080p, 31 at 2K and 63 at 4K. A 5-second 1080p clip costs 100 credits; a 16-second 4K clip costs 1,008. Image-to-video and reference-to-video cost the same, and reference images and voice clips aren't charged extra.
Not entirely. New accounts get 12 free credits, and the shortest Vidu Q4 clip (3 seconds at 540p) costs 24. To generate, you need a credit pack (from $9.90 for 640 credits) or a subscription (from $15/month).
Pick Vidu Q4 when you start from images: it takes up to 15 references and voice clips, focuses on character acting and goes up to 4K. Pick MiniMax H3 when you only have a text prompt, or want reference videos and always-on stereo sound at 2K.
ShengShu offers Vidu Q4 Preview through its own Vidu API platform. If you'd rather not set up a separate account, Visiojoy's MCP server lets Claude, Cursor and other MCP clients animate an image with Vidu Q4, using the same credits as this website; reference-to-video is available here on the website.
No. Visiojoy is an independent AI video platform and is not affiliated with ShengShu Technology or Vidu. It gives you Vidu Q4 alongside Seedance, MiniMax, Wan, Grok Imagine and other models in one account, with one credit balance.
Yes, once you've made a purchase. Videos created after your first purchase (any subscription or credit pack) can be used commercially; videos created before your first purchase are licensed for personal, non-commercial use only. Make sure you have the rights to the faces, voices and products in your references.
Upload a photo or a few references, write the scene and generate a Vidu Q4 clip with dialogue and sound — all in the browser.
