Skip to main content

Uncensored — This model declares no content filter. SpicyAPI forwards the request without adding a platform content scan; legal and consent duties remain with the customer. Image to video — Uncensored; Video extension — Uncensored; Reference to video — Light.Browse every model at this filtering level
vidu/q4-preview/image-to-video

Vidu Q4 Preview image-to-video animates an opening frame into a 3 to 16 second clip at 540p up to 4k, keeping the shape of your frame. Sound and quoted dialogue are generated with the picture. In our tests NSFW content came through; since 9 October, the Very strong level is refused.

Filtering differs by endpoint: Image to video — Uncensored, Video extension — Uncensored, Reference to video — Light.

From $0.045/s
Run model

Workbench

Input

Image to video
302/20000
Required, up to 20,000 characters. Describe the movement, the camera and the sound after the frame. Put spoken lines in double quotes. Does not change the price. at least 1 characters · at most 20000 characters
1/1

Drop files or media links, paste, or browse · 1 file · Max 10 MB · JPG · PNG · WEBP · GIF

  • 1
1 / 1 files
Required. The opening frame: it fixes the people, look and shape of the clip. Use a public, directly downloadable link.
affects price
540p, 720p (the default), 1080p, 2k or 4k. This sets the price per second: 2k costs more than 1080p, and 4k about twice 2k. 2k and 4k return 10-bit HEVC files that some browsers cannot play inline. The shape follows your frame.
Durationaffects price
Whole seconds, 3 to 16, default 5. Billed per second together with the resolution. no lower than 3 · no higher than 16
Generate audio
On by default: dialogue, effects and ambience are generated with the picture. Turn it off for a clip with no audio track. Does not change the price.
Prompt expansion
Off by default. When on, the model rewrites and expands your prompt. In one run it toned a request down, so keep it off for literal direction. Does not change the price.
Seed
Keeps runs comparable while you change one thing at a time. It does not promise a frame-identical clip. Does not change the price. no lower than 0 · no higher than 2147483647
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
Example result
Extend
Video upscaling
Video editing

Related models and next tasks

Missing the model you need?

Tell us what to add and where you would use it. Requests go to the people who decide what gets onboarded next.

Turn an opening frame or up to 15 references into a clip that speaks its lines

Vidu Q4 Preview is the public preview of Vidu's Q4 video model, released on 7 October 2026. It has two endpoints: image-to-video animates an opening frame, and reference-to-video builds one scene from 1 to 15 images of people, outfits, products or places, with up to 3 voice clips. Clips run 3 to 16 seconds at 540p up to 4k, with dialogue, effects and ambience generated with the picture, and a line written in double quotes is spoken. In our tests image-to-video delivered the Mild, Moderate and Maximum levels of our five-level content test and softened Strong, but since 9 October it refuses Very strong; reference-to-video is lightly filtered and refuses or tones down the stronger levels.

Open the playground
Turn an opening frame or up to 15 references into a clip that speaks its lines
Model
Vidu Q4 Preview by Vidu (ShengShu Technology), a public preview released on 7 October 2026
Endpoints
Image to video from one opening frame; reference to video from 1 to 15 images and up to 3 voice clips; no text-to-video
Clip length
3 to 16 whole seconds at 24 fps, default 5
Resolution
540p, 720p (default), 1080p, 2k or 4k; 2k and 4k come back as 10-bit HEVC files
Sound
Dialogue, effects and ambience on by default, one switch turns them off; lines in double quotes are spoken
Content filter
Image to video: Moderate and Maximum levels delivered, Very strong refused since 9 October. Reference to video: lightly filtered, refuses or tones down the stronger levels
Image to video
Your frame sets the look, the prompt sets the actionThe clip starts from your opening frame and keeps its shape; there is no aspect-ratio field. In our runs the people, wardrobe and light of the frame carried through, so the prompt only has to say what happens next, what the camera does and what is said.
Reference to video
Up to 15 references in one sceneSend people, outfits, products and places as separate images and name each by its position in the prompt. Three references, a portrait, a silk garment and an empty fitting room, came back as one scene, and two portraits kept both faces recognisable.
Content
Two endpoints, two different filtersOn image to video, the Moderate and Maximum levels came through and the Strong level came back softened by the framing; the Very strong level stayed implied on 8 October and has been refused since 9 October. Reference to video refuses or tones down the stronger levels. Refused requests are not charged.

Working with Vidu Q4 Preview

Direct the next five seconds, line included

Direct the next five seconds, line included

Our opening frame was a woman in a black satin slip dress at a three-way mirror in a velvet dressing room. The prompt listed five beats in order, with one spoken line, "Help me with this?", near the end. The 1080p clip played the beats and the line was spoken. The top-left panel is the opening frame, the other three are moments from the clip.

Name each reference by its position

Name each reference by its position

Reference to video takes 1 to 15 images. We sent a portrait of a woman, a product photo of an emerald silk garment and an empty boutique fitting room, and named each by its position in the prompt, starting with "the woman from the first image". Her face, the garment's colour and sheen, the gilded mirror and the pink velvet curtains all came back in one 720p scene.

  • A voice clip sets the voice, the prompt sets the words

    Reference to video also takes up to 3 audio clips; Vidu asks for MP3, 3 to 12 seconds each. The clip shapes the voice but supplies no words: a run that asked for the words of the reference audio without writing the line came back with no speech at all. Write the line in double quotes and say it is spoken in the voice from the reference audio.

  • Draft at 540p, finish at 1080p

    Image to video took about 2 minutes of model time for a 5-second clip at 540p or 720p in our tests, and reference to video about 3 to 10 minutes. Try the prompt at 540p with a fixed seed, then rerun the version you like at 1080p. 540p, 720p and 1080p come back as H.264 MP4 files that play almost anywhere.

  • Treat 2k and 4k as master files

    At 2k and 4k the clip is a 10-bit HEVC (H.265) MP4. The files are large, a 3-second 4k clip was 8.6 MB, and some browsers cannot play them inline, so open them in a desktop player or editor, or transcode them. The price per second follows the resolution tier: 2k costs more than 1080p, and 4k about twice 2k.

Who uses Vidu Q4 Preview

  • Characters who say their lines

    Storytellers and short-drama makers

    Characters who say their lines

    Send a portrait and a short voice clip, write the line in quotes and pick 9:16 for a phone screen. Our woman with long black hair and red lipstick turned a key in a candlelit doorway and whispered "Lock the door. Nobody is coming up tonight... and I am not in any hurry." over nine seconds, in the voice from the clip. The left panel is the reference portrait.

  • Vertical clips that keep your frame

    Social and short-form teams

    Vertical clips that keep your frame

    Image to video keeps the shape of the opening frame, so a 9:16 still gave a 720 x 1280 clip at 720p with no extra setting. In our test a man rose from a marble bench and walked toward the camera through drifting steam, with steam and dripping water in the sound. The left panel is the opening frame.

  • Start from your users' own frames

    Uncensored platform builders

    Start from your users' own frames

    Image to video delivered the Mild, Moderate and Maximum levels of our content test, so an unfiltered app can animate the frames its users bring. The Strong level came back softened by the framing, and the Very strong level has been refused since 9 October. Everyone on screen must be an adult.

  • Two people from two portraits

    Series and character creators

    Two people from two portraits

    Send one portrait per person and refer to each by position. Two portraits of two men in suits came back in a mirrored brass hotel elevator, the bearded one loosening the other's tie, with both faces recognisable. Keep a two-person reference shot to one slow beat, and reuse the same portraits in the same order for every shot of a sequence.

From first frame to final cut with Vidu Q4 Preview

  1. Pick the endpoint by what you have

    An opening frame goes to image to video; separate photos of people, outfits, products or places go to reference to video. For the Moderate and Maximum levels, start from a frame of the people on image to video.

  2. Write beats, camera and sound

    Describe what happens after the first moment in order, say what the camera does, and put every spoken line in double quotes. Keep prompt expansion off so your wording is used as written.

  3. Draft at 540p with a fixed seed

    Run a few versions at 540p and 3 to 5 seconds, and check faces, hands and the spoken line before you go longer.

  4. Finish at 1080p

    Rerun the winner at 1080p with the same prompt and seed. Choose 2k or 4k only if your editor handles 10-bit HEVC files.

  5. Plan for refusals and silence

    A refused request fails after about 6 to 8 minutes and is not charged; move a Moderate- or Maximum-level scene from reference to video to image to video. Turn sound off when you will add your own soundtrack; the price stays the same.

Vidu Q4 Preview prompts worth copying

  • An opening frame that speaks

    For image to video from a frame of one adult. List the beats in order and quote the line.

    The adult woman in the black satin evening dress turns from the dressing-room mirror toward the camera, tucks a loose curl behind her ear and smiles. She says, "You're late. Sit down, the show starts in five minutes." Warm vanity bulbs glow behind her, the camera pushes in slowly, soft room tone.

  • Three references, one fitting

    For reference to video with a portrait, a garment photo and a location photo, in that order. Name each image by its position.

    The adult woman from the first image stands in the boutique from the third image, wearing the long silk coat from the second image over a black turtleneck. She fastens the belt, turns to the mirror, looks back at the camera with a confident smile and says, "This one. I'll wear it out." Soft lamp light, slow dolly-in, the rustle of silk.

  • A portrait and a voice clip

    For reference to video with one portrait and one MP3 voice clip of 3 to 12 seconds. The words must be in the prompt.

    The adult woman from the reference image, in a dark green velvet dress, stands by a rain-streaked window in a candlelit study. She turns to the camera with a small, knowing smile and says in the voice from the reference audio: "You kept me waiting. Tell me everything." Close-up, shallow depth of field, flickering candlelight, rain on the glass.

More Vidu Q4 Preview questions

Use it when the scene needs a voice

Start from an opening frame when you already have the shot, or from references when people, outfits, products and places must match.

Write the beats in order, quote every line, draft at 540p and finish at 1080p. For the Moderate and Maximum levels, use image to video from a frame of the people; reference to video refuses or tones them down. Image to video has refused the Very strong level since 9 October, and refusals are not charged.

All Vidu models5