seedinteger
Random seed (0-2147483647) for reproducible results. Omit for a random seed; the same seed may still produce slightly different results.
Minimum: 0
Maximum: 2147483647
alibaba/wan/3.0/image-to-videoWan 3.0 by Alibaba - animate still images into dynamic videos with AI. Transform photos into cinematic clips with natural motion, camera movement, and optional audio generation.
Try the model
PNG, JPEG, WebP, or GIF · 20 MiB maximum
Customize your input with more control.
PNG, JPEG, WebP, or GIF · 20 MiB maximum
What would you like to do next?
Select an example to load its inputs and result into the playground.
{
"duration": 20,
"image": "https://static.sandbase.ai/examples/d2093d34-17ff-42a9-8b36-45e0c52f8028/input/0e8430dd-0e38-49d9-b437-96faa0d0bbc1.png",
"prompt": "(0:00 - 0:03) Camera: Medium shot. Scene: A ink-wash bamboo forest scene from Image 1. A swordsman wearing a bamboo hat and dressed in a dark robe, his silhouette standing in the center surrounded by thick fog and swaying bamboo shadows. Action: The swordsman's left hand lightly rests on the sword hilt, bamboo leaves fall with the wind. Atmosphere: Silent, oppressive, with clearly visible ink-wash grain texture. (0:03 - 0:08) Plot introduction. Camera: Close-up. Plot: The swordsman senses killing intent. Scene: The camera rapidly pushes in to a close-up beneath the swordsman's bamboo hat. Detail: Beneath the hat brim, a pair of sharp, resolute eyes appear, gaze like a blade. Ink-wash \"sweat drops\" slide down from the forehead. Action: The swordsman's right hand slowly grips the sword hilt, knuckles turning white. The bamboo leaves fall faster. (0:08 - 0:13) Camera: Fast orbit. Scene: The camera rapidly orbits around the swordsman. Plot: Two assassins, also ink-wash silhouettes, burst out from deep in the bamboo forest, wielding twin hooks and short blades. Detail: The assassins' movements are also composed of ink-wash brushstrokes, flowing like ink traces. The camera sweeps past the assassins' figures rapidly, emphasizing speed. (0:13 - 0:18) Camera: Handheld feel, high-speed capture. Combat: The swordsman draws his blade sharply — the blade light is a pure white ink-wash crack. The assassin attacks with twin hooks. Detail: Strike one — the swordsman spins and slashes, blade colliding with the assassin's twin hooks, sending sparks of black and white with calligraphic flair and ink drops flying. Strike two — the swordsman feints, uses a bamboo pole to catapult himself, and strikes down from above. The assassin rolls to dodge. The bamboo pole shatters into scattered ink traces. Strike three — the swordsman's blade tip points at the other assassin's throat; the assassin blocks with a short blade, and the camera captures the fine detail of ink traces cracking where the blades grind. (0:18 - 0:20) Camera: Slow motion to freeze frame. * Scene: The swordsman's blade stops at the assassin's neck, the assassin frozen. * Detail: The swordsman's bamboo hat is slightly tilted from the fight. The bamboo forest returns to calm, more bamboo leaves dancing. * Atmosphere: The camera slowly pulls back to medium shot, freezing on the swordsman sheathing his blade and the assassin falling. * Text (optional): Ink-wash style text appears at the bottom: \"The world of martial arts — settled with a single stroke.\" Style note: Throughout, maintain the highly black-and-white, ink-wash painting style of Image 1 . Actions must have a calligraphic brush feel, speed must be fast, details must be rich (ink drops, sparks, ink crack marks)."
}{
"duration": 15,
"end_image": "https://static.sandbase.ai/examples/5ab85ae0-70bc-4a81-b979-5d6c3bc7c3d7/input/24945908-0fe1-497b-87ef-a903428d31c8.png",
"image": "https://static.sandbase.ai/examples/5ab85ae0-70bc-4a81-b979-5d6c3bc7c3d7/input/490e83b7-53eb-4466-828b-9e44431e4f8e.png",
"prompt": "[One continuous take, slow aesthetic camera movement, no cuts, 15-second long shot, rock-color mineral pigment texture colliding with 3D rendering, cyan-blue × gilt gold contrasting colors] Global setting: Chinese-style rock-color painting texture, mineral pigment granular texture, gold foil texture, strong cyan-blue and gilt gold contrasting colors. The scene is West Lake at dawn in thin mist: the lake surface shows matte cyan-blue ink wash, golden morning light pierces through clouds forming Tyndall light columns, the three stone pagodas of Three Pools Mirroring the Moon and Leifeng Pagoda are faintly visible in the mist, with lotus leaves and lotus flowers in the foreground. The main subject is the West Lake water goddess \"Xizi\": rock-color painting texture, oval face, willow-leaf eyebrows, phoenix eyes, a cinnabar floral ornament between her brows, skin like warm jade emitting a soft glow, a pale golden halo floating behind her head; hair swept up high with a gilt lotus hairpin, her skirt and shawl composed of flowing lake water, cyan-blue gradient, with shimmering water reflections on the surface. Audio and sound effects: Background features gentle guzheng and bamboo flute, with ambient sounds of subtle water ripples and morning birdsong. Starting at second 8, an ethereal, gentle female voice slowly recites: \"If West Lake were compared to the beauty Xizi—\"; during the turn-back segment, she continues: \"Whether lightly or heavily adorned, she is always beautiful.\" The pace is slow, with slight reverberation and lingering resonance. [0-3s: A water drop] Extreme macro close-up, at the edge of a lotus leaf at dawn, a water droplet gathers to fullness, reflecting golden morning light, slowly falling. The camera follows the droplet with a tilt down, the droplet falls into the lake, creating rings of gold-blue ripples that spread outward layer by layer. [3-7s: Water coalesces into form] At the center of the ripples, lake water slowly rises and coalesces into the form of the goddess \"Xizi\": the lake water transforms into her cyan-blue skirt and semi-transparent shawl, with fine water streams circling around her. She slowly rises from the lake center, eyes gently closed then slowly opening, the gilt lotus hairpin glowing faintly. The camera slowly orbits her half-lowered gaze and the skirt woven from flowing water. [7-11s: A lotus with every step] She walks barefoot on the lake surface, each step creating rings of golden ripples, with lotus flowers blooming one after another within the ripples. The camera pulls back to a medium-wide shot: the mist disperses, revealing the three stone pagodas of Three Pools Mirroring the Moon glowing with warm light in the distance, Leifeng Pagoda appearing in the morning light, and weeping willows brushing the water along the shore. At second 8, the female voiceover begins: \"If West Lake were compared to the beauty Xizi—\" Her shawl trails a long gold-blue water-light behind her. [11-15s: The turn — always beautiful] She stops, turns to look at the camera, the camera slowly pushes in to a medium close-up: morning light falls on her face, the lake-water shawl gently drifts at her side, a few glowing water droplets fall from her hair tips, transforming into fine light particles in the air. The voiceover continues: \"Whether lightly or heavily adorned, she is always beautiful.\" She gives a gentle smile, with Leifeng Pagoda and the hazy Three Pools Mirroring the Moon behind her. In the final second, the frame completely freezes on her turning gaze, composed as a perfect cinematic poster."
}Human preference
Elo ratings from blind side-by-side votes, compared with other image and video models.
Rankings sourced from Artificial Analysis Arena
Image-to-video arena (silent): people pick the better of two clips animated from the same image, without seeing the model names; audio is not judged.
Start a clip from your own picture. image becomes the exact first frame; add end_image and the model interpolates the motion between the two. Clips run 2–30 seconds with native audio.
| Field | Type | Default | Notes |
|---|---|---|---|
prompt | string, required | — | Up to 5,000 characters. |
image | image URL, required | — | First frame. JPEG/PNG/BMP/WebP, 240–8000px per side, ≤20MB, no transparency. |
end_image | image URL | — | Last frame; needs image. |
duration | 2–30 or auto | 5 | auto lets the model pick the length. |
resolution | 480p / 720p / 1080p | 720p | Sets the per-second rate. |
aspect_ratio | adaptive / 21:9 / 16:9 / 4:3 / 1:1 / 3:4 / 9:16 | adaptive | adaptive follows the prompt and inputs. |
generate_audio | boolean | true | Native dialogue, music, and effects. Same price on or off. |
seed | integer 0–2147483647 | random | Fix it to compare prompt changes. |
You pay for output seconds only: $0.05 at 480p, $0.10 at 720p, $0.20 at 1080p. Input images, audio, and video are not metered. The frame images are free.
| Output | Billed | 480p | 720p | 1080p |
|---|---|---|---|---|
| 2s | 2s | $0.10 | $0.20 | $0.40 |
| 5s (default) | 5s | $0.25 | $0.50 | $1.00 |
| 10s | 10s | $0.50 | $1.00 | $2.00 |
| 30s | 30s | $1.50 | $3.00 | $6.00 |
With duration: "auto" the run is pre-authorized at 30 seconds, then settled on the length that was actually generated. Failed runs are not charged. Output files are copied to SandBase storage and the returned URL stays valid for 7 days; download anything you need to keep.
aspect_ratio on adaptive to keep the image's own proportions.The subject slowly turns toward the camera as a breeze moves the fabric; subtle handheld push-in, natural daylight, no dialogue.
Other Wan 3.0 endpoints: alibaba/wan/3.0/video (all modes in one schema), alibaba/wan/3.0/text-to-video, alibaba/wan/3.0/reference-to-video, alibaba/wan/3.0/video-edit, alibaba/wan/3.0/video-extend.
Start building
Submit this model through the SandBase Unified Run API.
# The quoted heredoc keeps Unicode and shell metacharacters unchanged.
result=$(curl --fail-with-body --silent \
-X POST "https://api.sandbase.ai/v1/run" \
-H "Authorization: Bearer $SANDBASE_API_KEY" \
-H "Content-Type: application/json" \
--data-binary @- <<'SANDBASE_JSON'
{
"model": "alibaba/wan/3.0/image-to-video",
"image": "https://static.sandbase.ai/samples/image-to-video/i2v-black-cat-sky-img0.png",
"duration": 5,
"resolution": "720p",
"generate_audio": true
}
SANDBASE_JSON
)
run_id=$(printf '%s' "$result" | jq -r .id)
for attempt in $(seq 1 120); do
status=$(printf '%s' "$result" | jq -r .status)
case "$status" in completed|failed|timeout) break ;; esac
sleep 2
result=$(curl --fail-with-body --silent \
-H "Authorization: Bearer $SANDBASE_API_KEY" \
"https://api.sandbase.ai/v1/run/$run_id")
done
status=$(printf '%s' "$result" | jq -r .status)
[ "$status" = completed ] || { echo "Generation ended: $status" >&2; exit 1; }
printf '%s\n' "$result"Create a SandBase API key and set it in your server environment. All requests use Bearer authentication.
export SANDBASE_API_KEY="YOUR_API_KEY"Keep the key on your server. Do not expose it in browser code or commit it to a repository.
POST to /v1/run with this model ID and its inputs. The request example above includes submission and polling.
Use the returned run ID to poll until the status is completed, failed or timeout. Read output only after completed.
curl "https://api.sandbase.ai/v1/run/$RUN_ID" \
-H "Authorization: Bearer $SANDBASE_API_KEY"Pass a publicly accessible file URL in the model’s image, video or audio field. Supported formats and limits are shown beside each Playground input.
Use the upload or replace control on supported image fields in the Playground. Video URL fields accept a hosted driving video; they do not imply video-upload support.
seedintegerRandom seed (0-2147483647) for reproducible results. Omit for a random seed; the same seed may still produce slightly different results.
Minimum: 0
Maximum: 2147483647
imagestringrequiredURL of the image used as the first frame (JPEG, PNG, BMP, or WebP). Cannot be combined with `images`, `videos`, or `audios`.
promptstringrequiredText prompt describing the motion to generate.
durationvalueOutput duration in seconds (2-30), or `auto` to let the model pick a length from the prompt and inputs. `auto` is pre-authorized at 30 seconds and billed on the actual output length.
Default: 5
Possible values: "auto", 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
end_imagestringURL of the image used as the last frame (JPEG, PNG, BMP, or WebP). Cannot be combined with `images`, `videos`, or `audios`.
resolutionstringOutput video resolution tier.
Default: "720p"
Possible values: "480p", "720p", "1080p"
aspect_ratiostringOutput aspect ratio. Omit to match the input image.
Possible values: "21:9", "16:9", "4:3", "1:1", "3:4", "9:16"
generate_audiobooleanGenerate a synchronized audio track. Pricing is the same either way.
Default: true
{
"model": "alibaba/wan/3.0/image-to-video",
"image": "https://static.sandbase.ai/samples/image-to-video/i2v-black-cat-sky-img0.png",
"duration": 5,
"resolution": "720p",
"generate_audio": true
}A completed run exposes its result in output. The following is the model’s saved example, not a new generation.
[
{
"poster": "https://static.sandbase.ai/examples/posters/e58093b86ebca247b3e2.jpg",
"source": "Alibaba Cloud",
"source_url": "https://www.alibabacloud.com/help/en/model-studio/wan3-video-generation-guide",
"url": "https://static.sandbase.ai/examples/d2093d34-17ff-42a9-8b36-45e0c52f8028/output/bf97b9a9-1779-4920-86b8-34119ea0ce63.mp4"
}
]