Vidu Q3 Uncensored Test: Four Versions Compared

You have a still of an adult character and you want it to move. Vidu Q3 comes in four versions on SpicyAPI (Q3, Q3 Pro, Q3 Turbo and Q3 Spicy), all of them image-to-video and all labelled Uncensored (Vidu's own app and API can screen or block inputs and outputs; SpicyAPI adds no content filter). The practical question is which one returns the scene you wrote instead of a toned-down version of it. Based on our September 2026 tests, this page gives the results, the prices and the prompts to start with.

Content levels, each going further than the last: L1 Suggestive (lingerie, a slipping strap), L2 Mature themes, L3 Strong, L4 Intense (an intimate scene between two adults) and L5 Extreme (consensual adult play, and cinematic gore). ✅ means the clip came back as written, ⚠️ toned down (the call still succeeds and is billed), ❌ refused. Level definitions: review methodology.

The short answer

• Nothing was refused, and all four versions delivered L1, L3 and both L5 probes every run. The differences sit at L4 and L2. Q3, Q3 Pro and Q3 Turbo cleared L4 once in three and toned it down twice. Q3 Spicy cleared it three times in three, and it was the only version to slip on L2 (two of three).
• In the hands-on clips every version kept the content level of the still, but not always the picture. Pro relit the room and framed the clip in a black border, and the L5 prompts replaced a daylight bedroom with a red room.
• Image-to-video only. Every Vidu request starts from a still. For uncensored video from text alone, use MiniMax H3.
• Prices for 5 seconds at the default 720p: $0.33 on Turbo, $0.71 on Spicy, $0.75 on Q3. Pro is fixed at 540p ($0.27). Drop to 540p for $0.21 on Turbo and $0.33 on Spicy.
• Pick by level, not by price. Turbo is the cheap way to try motion on L1 to L3 stills. For an L4 scene start on Spicy at 540p, which costs the same as Turbo at 720p. Spicy has no seed, scored lowest in our blind quality test, and can go further than a mild prompt asks.

Try Vidu Q3 Spicy →

How far each Vidu Q3 version goes

Image-to-video at 540p, two seconds, audio off, three runs per rung on each version, judged on the frames. The Q3 Spicy test report lists the same results and adds one earlier run, which is why it says 19 runs.

No cell contains a refusal. Uncensored means we add no content filter, so the model decides what comes back. A toned-down clip still succeeds and is billed; the difference shows only in the frames, so play every clip to the last frame before you accept it.

Hands-on clips: what a prompt does to the picture

One clip per version from an L3 still (one adult woman on a white bed in daylight). None was refused or toned down; they show what a prompt does to the picture.

Q3 · Q3 Pro · Q3 Turbo · Q3 Spicy

The frames

The frames below are pixelated by us; the models returned them unblurred.

L3: Pro and Turbo on the same still and prompt. Both kept the level of the still. Pro followed the lighting cue, turned the daylight into a warm lamp glow and ended inside a black border. In our Turbo clip the cue was ignored: it kept the window daylight and lowered the camera while she opened her eyes and looked into the lens.

Left: the opening still we sent. Middle: Vidu Q3 Pro at 2.5 seconds. Right: Vidu Q3 Turbo at 1.5 seconds. L3, 4-second clips.

Pro's first second. A black panel covered the picture for the first 0.6 seconds, then the picture came back smaller inside a black frame and stayed that way.

Sound. On by default through the API (some Playground forms preselect it off), with a separate bgm switch also on; neither changes the price, and bgm has no effect on Q3 models.

How to write Vidu Q3 prompts

The still decides who is on screen and how far the frame goes; the prompt decides what happens next. On Q3 and Q3 Spicy the prompt can bring in a second person, props and a new light. Put the level in the frame and the action in the words: action, then camera, then light, then sound. Don't describe the picture again.

Stay on your still (Q3 Turbo draft)

She slowly opens her eyes and turns her head on the pillow to look into the lens, arches her back slightly and settles again. The camera lowers and pushes in slowly. Keep the daylight from the window. Quiet room tone.

Run this prompt on Q3 Turbo →

In our Turbo clip the daylight stayed even though the prompt asked for warm lamp light, which makes Turbo the safest draft tier when your frame is already right. To keep the light, say so as above; to change it, grade the still first or relight on Pro.

Relight the same still (Q3 Pro)

She arches her back and turns toward the camera on the bed. Warm low lamplight replaces the daylight. Slow push-in, then hold on her face. Soft breathing, no music.

Run this prompt on Q3 Pro →

Pro followed this kind of relight. Its clip opened with a black panel for 0.6 s and kept a black border; trim the start if it appears.

Add a second person (Q3 Spicy)

A second adult, a man, steps into frame behind her and rests his hands on her bare shoulders; she leans back against him and looks up at the camera. The daylight warms to a low amber lamp. Slow push-in, one continuous shot.

Run this prompt on Q3 Spicy →

Keep the same shape at the upper levels: name the new person, what they hold, what they do and what the light becomes. Plain words for props went through as written.

No prompt, or almost none (Q3 Spicy)

Spicy is the only Vidu version that takes a still with no prompt at all, which suits a folder of stills. It can show more than a mild prompt asks for, so check each clip. For a specific beat, keep the prompt this short:

Slow breath, she rolls onto her side toward the lens and pushes her hair back. Hold on her eyes.

Run this prompt on Q3 Spicy →

Camera, light and sound in one take (the booth clip)

She rolls the black stiletto off her pointed toes and it drops onto the velvet, then draws that leg down along the back of the booth and crosses it over the other. Her other hand pulls the fallen strap back up onto her bare shoulder and she holds the look into the lens, lifting her chin. The red neon on the right breathes brighter and the amber sconce flickers. Slow low tracking dolly moving in past the edge of the table, quiet bar room tone.

Run this prompt on Q3 Spicy →

Feet first, then her hand and her eyes, then the lights, then the camera and the room sound: one beat per sentence, in the order they happen.

Two beats joined by "then" (the bed-end clip)

She swings her crossed bare feet slowly back and forth in the lamp light and the fluffy fox tail sways with them. Then she pushes up onto her hands and knees and crawls one slow step toward the camera, hair falling forward, and looks straight down the lens. Warm bedside lamp, sheets shifting under her, soft handheld micro-drift, shallow depth of field.

Run this prompt on Q3 Turbo →

Two actions in order, joined by "then", and the camera gets its own clause. Naming a detail from the still (the fox tail) makes it move with the action.

Weak vs strong

Weak: "Beautiful, cinematic, stunning, 4k." Adjectives describe a picture that is already in your frame.

Strong: one action, one camera move, one light, one sound, in the order they happen.

Where Vidu trips up, and the fix

• A toned-down clip looks like a success. The job status gives no hint. Fix: play the whole clip, and if an L4 scene matters, run it on Spicy.
• The prompt can replace your set. L5 prompts on Q3 and Spicy swapped the daylight bedroom for a red room. Fix: say "same room, same daylight".
• Spicy can show more than you wrote. Fix: review before use, and say what stays in the frame.
• Pro's clip opened with a black panel and kept a black border; Turbo's clip ignored a relight cue. Fix: trim Pro's first second; to change the light, grade the still first or relight on Pro.
• Q3, Pro and Turbo reject a request with no prompt. It fails before anything runs and costs nothing. Fix: add one line, or use Spicy.
• There is no aspect-ratio setting. The still decides the shape; a square still came back as 720 × 720 at 540p and a 3:2 still as 1176 × 784 at 720p. Fix: crop the still to the shape you deliver. The still can be JPG, PNG, WEBP or GIF, up to 10 MB, with an aspect ratio between 1:4 and 4:1.
• The move comes out too small or too big. Fix: write its size into the prompt: "a half turn", "two slow steps toward the camera". The movement_amplitude setting has no effect on Q3 models.
• Spicy has no seed, so every re-roll is a new take. Fix: use Q3, Pro or Turbo, which accept a seed, for side-by-side comparisons.
• 16 seconds is the ceiling. Fix: send the last frame back as the next still (longer uncensored video guide).

Which Vidu Q3 version should you use?

The short answer: choose by how far the scene goes. For L1 to L3, draft on Turbo at 540p for $0.21 a clip and move the keeper to Q3 Spicy only if you need 1080p or a prompt that builds the scene. For L4, do not draft on Turbo: it toned down two of three L4 clips that Spicy delivered, so a Turbo draft cannot tell you how the Spicy version will look. Start on Spicy at 540p, which at $0.33 for 5 seconds costs what Turbo costs at 720p.

What Spicy trades away. It costs 5% less than Q3 at every resolution and cleared L4 every time, but it has no seed, and in our blind quality test (ordinary prompts, scored by three judges who did not see the model names) it scored 41.9 out of 100, against 67.3 for Q3, 67.7 for Turbo and 63.1 for Pro. Its weakest criteria were consistency over time (33.7, against 61.9 to 67.9 for the others) and prompt adherence (38.3, against 64.7 to 72.0). Review clips rather than run Spicy unattended. Our Q3 versus Q3 Spicy comparison has the field-by-field difference.

Pro or Turbo? Turbo for most work: it is cheaper per second, offers 720p, and completed a two-person motion that Pro barely animated. Pick Pro when you want a prompt to relight the scene at 540p; check the first second for a black panel.

Against the neighbours: MiniMax H3 for uncensored video straight from text, delivered L4 and L5 as written (5 seconds from $0.125, up to 15 seconds); Seedance 2.5 Spicy for one unbroken take up to 30 seconds or 4K (5 seconds from $1.08); Vidu for 1-second test clips and a Spicy build that turns one still into the scene you describe. More: MiniMax H3 uncensored test, Seedance versions tested.

What a clip costs

The API defaults to 720p on every version except Pro. The Playground forms can differ (Q3 Spicy preselects 1080p, Turbo 720p), so set the resolution before you run. Sound on or off costs the same. A request that fails its checks is rejected before it runs, free. Prices as of 3 October 2026, rounded to the cent; each model page shows the live estimate before you submit.

Run it: Playground first, then the API, or an Agent

In the Playground. Create an account, open a version's page and stay on the Playground tab in Form view.

1. Open Vidu Q3 Turbo for a draft, or Vidu Q3 Spicy for an L4 scene.
2. Add your still under Input image (JPG, PNG, WEBP or GIF, up to 10 MB) and paste one of the prompts above into Prompt. Spicy accepts an empty prompt; Q3, Pro and Turbo need one.
3. Set Resolution to 540p and Duration to 5 for a first look: $0.21 on Turbo and $0.33 on Spicy. The form preselects 720p on Turbo and 1080p on Spicy, so change it before you run. Duration and Resolution change the price; Generate audio and Background music do not.
4. Check the price shown for your settings, press Run model, and play the clip to the last frame. Send the keeper again at 1080p when the motion is right.

Then the API, once a prompt works and you want it in your own app. This one animates a still on Q3 Spicy with no prompt:

curl -X POST https://api.spicyapi.ai/api/v1/jobs/createTask \
  -H "Authorization: Bearer $SPICY_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  --data '{
    "model": "vidu/q3-spicy/image-to-video",
    "input": {
      "image_url": "https://cdn.example.com/your-still.png",
      "resolution": "540p",
      "duration_seconds": 5
    }
  }'

For Turbo, swap the model for vidu/q3-turbo/image-to-video and add a prompt. Fields, waiting for the result and errors are in the API reference and the tasks guide.

Or let an Agent run it. An agent can make the same call through the CLI, the MCP server or the Agent Skill. All three read the live model schema, and the CLI and MCP show the price and ask you to confirm before creating a task. With SPICY_API_KEY set:

# CLI
npx @spicyapi/cli tasks create --model vidu/q3-spicy/image-to-video \
  --input-json '{"image_url": "https://cdn.example.com/your-still.png", "resolution": "540p", "duration_seconds": 5}' --wait

# MCP server: the command your assistant's MCP config runs
npx --yes --package=@spicyapi/mcp spicyapi-mcp

# Agent Skill
npx skills add spicyapi-ai/spicy-skill

FAQ

Is Vidu uncensored?

On SpicyAPI, all four versions are labelled Uncensored and none refused a clip; the issue is toning down, which hit L4 on three versions. Vidu's own app and API are different: their terms let Vidu screen, edit or block any input or output and prohibit obscene content (Vidu API Terms of Use).

What's the difference between Vidu Q3 and Vidu Q3 Spicy?

Q3 Spicy is a LoRA-tuned build of Vidu Q3. It runs without a prompt, has no seed, costs 5% less per second than Q3, and cleared L4 in three of three runs against one of three for Q3. The trade-off is lower blind-test quality.

Vidu Q3 Pro or Turbo for uncensored video?

Turbo for most work: it is cheaper per second, offers 720p and stays closest to your still. Pick Pro only when you want the prompt to relight the scene at 540p, and check its first second for a black panel.

Can a prompt take a tamer still further on Vidu?

Partly. On Q3 and Q3 Spicy a prompt brought in a second person, props and a new light, but the still sets the baseline level.

Can Vidu Q3 do text-to-video?

Not on SpicyAPI: every Vidu version here is image-to-video. Make the still with an uncensored image model, or use MiniMax H3 text-to-video.

How long can a clip be, and how long does it take?

1 to 16 whole seconds on all four. A 1-second clip at 540p came back in about 35 seconds.

Can I use a photo of a real person?

Only with that person's consent.

Your part

We don't filter inputs, so the checks are yours. Everyone on screen must be an adult. Any real, identifiable person must have consented. You need the rights to every still you upload, and you're responsible for the law where you publish.