Turn a still into a scene. Vidu Q4 Preview animates your images with dialogue, sound and its own camera cuts.
Add an image, write what happens, pick a length. The exact cost is on the button before you render, and a failed video costs nothing.
Length
05sec
Resolution
Every clip below was made on this site with the exact prompt shown. Play one and watch the script light up line by line, then load it into the editor.
Interrogation
Image to Video · 720p · 8 s
A shouted line, real anger and a camera that pushes in to the eyes. Lip-synced, with sound, from one photo.
INT. INTERROGATION ROOM - NIGHTIMG 8s
The detective slams both hands on the table and the swinging lamp throws moving shadows across his face. He shouts: "Where were you that night?!" The camera pushes in fast to an extreme close-up of his eyes. The suspect flinches. The bang of the table, the creak of the swinging lamp, heavy breathing.
Two characters and a location from separate images: a standoff, a line of dialogue, then the clash. Faces stay consistent across the cut.
EXT. STONE BRIDGE - RAINREF 10s
In 
@3, 
@1 and 
@2 face each other in the middle of the bridge in the pouring rain, lanterns swaying. @2 raises his sword and says: "Only one of us leaves this bridge." @1 draws his katana. They charge and their blades clash in a burst of sparks. Cut to a slow-motion close-up of the crossed blades with rain bouncing off the steel. Thunder, rain, the ring of steel on steel.
Animals and action in one shot: the panther leaps, lanterns burst into sparks, and the camera whips around to follow.
EXT. NIGHT MARKET - RAINIMG 8s
The black panther leaps off the stall roof and crashes through the lanterns, which burst into showers of sparks. The man glances back in terror and sprints faster, splashing through puddles as stalls topple behind him. The camera whips around to follow the panther, then cuts to a low close-up of its snarling face as it lands. Heavy rain, a deep animal roar, crashing wood, people screaming.
Frozen rain, a 360° orbit and a shockwave: effects that react to the light and the space around them.
EXT. CITY STREET - NIGHTIMG 8s
She snaps her fingers and every raindrop freezes in mid-air, glittering in the lamplight. The camera makes a slow 360-degree orbit around her through the frozen rain. She sweeps her hand outward and all the drops blast away from her in a shockwave, shattering the puddles. A sharp finger snap, sudden silence with a low hum, then a deep whoosh and a crash of water.
Turn a packshot into a slow-motion ad: a crown of water, light streaks and an orbiting camera.
The sneaker slams down into the puddle in slow motion, throwing up a perfect crown of water that catches the amber light. Light streaks race past in the background as the camera orbits around the shoe. A deep bass hit, the splash, then a whoosh.
Fights and stunts stay fluid: a spin, a block, sparks from the lamp and a hard cut to a close-up.
She spins to block a strike from off screen and slides back across the wet roof. The lamp above her bursts into a shower of sparks. The camera whips around to follow her, then cuts to a low-angle close-up as she exhales. Heavy rain, a roll of thunder, sharp impacts.
One picture for Image to Video: it becomes the first frame. Or up to 12 reference images of characters, outfits, products and places for Reference to Video.
INT. INTERROGATION ROOM - NIGHT
The detective slams both hands on the table and the swinging lamp throws moving shadows across his face. He shouts: "Where were you that night?!" The camera pushes in fast to an extreme close-up of his eyes. The suspect flinches. The bang of the table, the creak of the swinging lamp, heavy breathing.
Say who does what, what they say and how the camera moves. In Reference to Video, call your images @1, @2… Add up to three MP3 voice clips to keep a character’s voice.
Pick a length from 3 to 16 seconds and a resolution up to 4K. Vidu Q4 renders the video with sound, and you download an MP4 from My videos.
Vidu Q4 Preview has two ways in. Image to Video brings one picture to life; Reference to Video builds a new scene around the people and things in your images. Both cost the same per second.
1 image = first frame@1@2@3Text-to-video is not part of Vidu Q4 Preview. Every video starts from an image.
Buy credits once and see the exact cost on the button before every video. Credits last 12 months, nothing renews, and failed videos are refunded automatically.
Try Vidu Q4 for real
$2.99one-time
Starter
Try a few videos
Creator
For a weekend project
Pro
For regular creators
Studio
For client and ad work
Vidu Q4 is the next flagship video model from Vidu, the AI video platform of Beijing-based ShengShu Technology. Its first public release, Vidu Q4 Preview, launched on October 7, 2026, ahead of the full Q4 model, so creators can use it on real projects while the final version is still being shaped.
Where earlier Vidu models focused on smooth motion, Vidu Q4 is built for performances: facial expression, emotion, body movement and voice are generated together, and the model plans its own camera moves and cuts across a shot. Audio comes out with the picture: dialogue and sound effects, not a silent clip you have to score later.
Slatecraft is an independent website that runs Vidu Q4 Preview through Vidu’s official API. You pay only for the seconds you generate, see the cost before you press Generate, and get your credits back automatically if a video fails.
Vidu Q3 added text-to-video and 16-second clips with sound. Vidu Q4 Preview keeps the 16-second length and the sound, raises the ceiling to 2K and 4K output with 10-bit color, and takes far more reference material: up to 15 images and 3 voice clips in one request, so a multi-character scene can stay consistent from shot to shot.
The trade-off is that Q4 Preview starts from images. There is no text-to-video mode in this release; if you only have an idea, create or pick a reference image first.
Describe the performance, not just the scene: who moves, what they feel, the exact line they say and how the camera follows. “@1 leans across the table and whispers ‘we’re out of time’; slow push-in, tense” works better than “two people talking”.
Keep one action per shot for short clips, and name the camera move. For dialogue, write the line in quotes and leave sound on; for a clip you will score yourself, use Reference to Video with sound off.
Vidu Q4 Preview is the first public release of Vidu Q4, the next-generation video model from ShengShu Technology’s Vidu, launched on October 7, 2026. It generates 3–16 second videos with synchronized dialogue and sound effects from one image (Image to Video) or from up to 15 reference images and 3 voice clips (Reference to Video), at up to 4K resolution.
Not in the Preview. Vidu Q4 Preview offers Image to Video and Reference to Video only, so every video starts from at least one image. You can still steer everything else (motion, dialogue, camera) with the prompt.
No. Every video runs on Vidu’s paid API. The cheapest way to try it is the trial pack: 300 credits, enough for 3 five-second 720p videos or 6 at 540p. Packs are one-time, nothing renews, and failed videos are refunded automatically.
No. Slatecraft is an independent service and is not affiliated with or endorsed by Vidu or ShengShu Technology. We run Vidu Q4 Preview through Vidu’s official API, with pay-per-second credits instead of a subscription.
You pay per second of video: 10 credits at 540p, 20 at 720p, 25 at 1080p, 40 at 2K and 80 at 4K. A 5-second 720p video costs 100 credits. The exact cost is on the Generate button before you start.
Usually a few minutes; longer clips and 2K/4K take longer. You can leave the page: finished videos are saved to My videos in your account.
Yes. With Sound on, Vidu Q4 generates dialogue and sound effects together with the picture. Put the line you want spoken in quotes in your prompt. In Reference to Video you can add up to three MP3 voice clips (3–12 seconds each) to keep a character’s voice consistent. For a silent clip, use Reference to Video and switch Sound off; Image to Video always renders sound.
Clear, well-lit JPG, PNG or WebP images up to 20 MB, no more elongated than 4:1. For Image to Video the picture is the first frame, so frame it the way the shot should start. For Reference to Video, one subject per image works best: a face, an outfit, a product, a location.
Finished videos are saved to My videos for at least 90 days. Download the ones you want to keep. Your uploaded images and voice clips are stored privately to run the generation. Videos are never published. To delete your files or your account sooner, email support@slatecraft.si from your account address.
Sexual content, anything involving minors, deepfakes or impersonation of real people, and images you don’t have the rights to. Our filters and Vidu’s moderation block these requests, and blocked videos are refunded automatically.
You can use the videos you generate from images you have the rights to, under our Terms of Service. Don’t upload photos of people without their consent, and don’t remove AI-generated labels where the law requires them.
Add a picture, write what happens, and hear your scene come to life in a few minutes.
