Omni ReferenceAlibaba
Wan 3.0 Prime
alibaba/wan-3.0-prime/video
Supports multimodal references including multiple images, video, and audio to generate videos with rich motion and synchronized audio-visual effects. Accurately preserves the appearance, motion characteristics, and visual style of referenced subjects, while supporting flexible combinations of references for versatile content creation and shot control.
Examples
Parameters
| Name | Type | Default | Constraints | Description |
|---|---|---|---|---|
| promptPrompt | textarea | — | ≤ 5000 chars | Up to 5000 characters. Either prompt or reference media is required. |
| imagesReference Images | image_upload_group | — | image/jpeg,image/png,image/bmp,image/webp · 0–10 items | Up to 10 images, ≤20MB each, 240–8000px per side, aspect ratio within 8:1. PNG alpha not supported. Cannot be combined with first/last frame. |
| audiosReference Audio | audio_upload_group | — | audio/wav,audio/mpeg · 0–5 items | Up to 5 clips, 1–15s each, 15s total, <15MB each. WAV and MP3. Cannot be combined with first/last frame. |
| first_frameFirst Frame | image_upload | — | image/jpeg,image/png,image/bmp,image/webp | Pins the first frame of the video. Cannot be combined with reference images, videos, or audio. |
| last_frameLast Frame | image_upload | — | image/jpeg,image/png,image/bmp,image/webp | Pins the last frame of the video. Cannot be combined with reference images, videos, or audio. |
| resolutionResolution | select | 1080P | 480P | 720P | 1080P | |
| aspect_ratioAspect Ratio | select | adaptive | adaptive | 16:9 | 4:3 | 1:1 | 3:4 | 9:16 | Adaptive picks a ratio from the prompt and input media. |
| durationDuration (seconds) | number | 5 | -1 ~ 30 · step 1 | 2–30 seconds, or -1 for smart duration decided by the model. With video input, input plus output must not exceed 30 seconds. Smart duration is pre-charged at 30 seconds. |
| generate_audioGenerate Audio | boolean | true | — | When off, the output video has no audio track. Pricing is unchanged either way. |
| enable_thinkingThinking Mode | boolean | false | — | Required for parsing documents, web pages, and complex image content. Not recommended without a document or link. |
| fileReference Document | file_upload | — | .docx,.doc,.xlsx,.xls,.pptx,.ppt,.pdf,.txt,.key,.pages,.numbers,.md · 0–1 items | One file max, ≤100MB, ≤50 pages. Requires thinking mode and cannot be combined with a web link. |
| linkReference Web Page | text | — | — | Public pages only, no login required. Requires thinking mode and cannot be combined with a document. |
| seedSeed | number | — | 0 ~ 2147483647 · step 1 |
API
Call this model through one unified REST API. Get a key on the API Keys page.
cURL
# 1) Submit — returns { "task_uuid": "..." }
curl -X POST "https://www.namifusion.com/api/v1/marketplace/run/alibaba/wan-3.0-prime/video" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"input": {
"resolution": "1080P",
"aspect_ratio": "adaptive",
"duration": 5,
"generate_audio": true,
"enable_thinking": false
}
}'
# 2) Poll until status is "completed", then read the output URLs
curl "https://www.namifusion.com/api/v1/marketplace/run/tasks/TASK_UUID" \
-H "Authorization: Bearer YOUR_API_KEY"Python
import time, requests
API_KEY = "YOUR_API_KEY"
HEADERS = {"Authorization": f"Bearer {API_KEY}"}
# 1) Submit
resp = requests.post(
"https://www.namifusion.com/api/v1/marketplace/run/alibaba/wan-3.0-prime/video",
headers=HEADERS,
json={
"input": {
"resolution": "1080P",
"aspect_ratio": "adaptive",
"duration": 5,
"generate_audio": True,
"enable_thinking": False
}
},
)
resp.raise_for_status() # 401/402/429/5xx stop here instead of polling a bad task
task = resp.json()
# 2) Poll until a terminal state (completed / failed / cancelled).
# This model is allowed up to 300s server-side.
deadline = time.time() + 360
while task.get("status") not in ("completed", "failed", "cancelled"):
if time.time() > deadline:
raise TimeoutError(f"still {task.get('status')} — keep the task_uuid and poll later")
time.sleep(3)
poll = requests.get(f"https://www.namifusion.com/api/v1/marketplace/run/tasks/{task['task_uuid']}", headers=HEADERS)
poll.raise_for_status()
task = poll.json()
print(task["status"], task.get("output"))JavaScript
const API_KEY = "YOUR_API_KEY";
const HEADERS = { Authorization: `Bearer ${API_KEY}` };
// 1) Submit
const resp = await fetch("https://www.namifusion.com/api/v1/marketplace/run/alibaba/wan-3.0-prime/video", {
method: "POST",
headers: { ...HEADERS, "Content-Type": "application/json" },
body: JSON.stringify({
"input": {
"resolution": "1080P",
"aspect_ratio": "adaptive",
"duration": 5,
"generate_audio": true,
"enable_thinking": false
}
}),
});
if (!resp.ok) throw new Error(`submit failed: ${resp.status} ${await resp.text()}`);
let task = await resp.json();
// 2) Poll until a terminal state (completed / failed / cancelled).
// This model is allowed up to 300s server-side.
const deadline = Date.now() + 360 * 1000;
while (!["completed", "failed", "cancelled"].includes(task.status)) {
if (Date.now() > deadline) throw new Error(`still ${task.status} — keep the task_uuid and poll later`);
await new Promise((r) => setTimeout(r, 3000));
const poll = await fetch(`https://www.namifusion.com/api/v1/marketplace/run/tasks/${task.task_uuid}`, { headers: HEADERS });
if (!poll.ok) throw new Error(`poll failed: ${poll.status}`);
task = await poll.json();
}
console.log(task.status, task.output);Documentation
Wan 3.0 Video Prime
One model for every input — text, images, video, audio, and documents — generating up to 30 seconds in a single shot.
Wan 3.0 Video is an all-in-one video generation model that upgrades duration, universal reference, and realism together. A single request produces up to 30 seconds of continuous video, giving narrative pacing room to breathe and letting complex camera language — continuous movement, one-take shots — play out fully. Beyond text, image, audio, and video, it is the first in the series to accept documents (doc, xls, ppt, pdf, md and more), turning office material directly into video.
🚀 Key Features
- Native 30-second generation: Produce a complete story rather than a single shot. Smart duration lets the model recommend the right length from your prompt.
- Omni reference across six modalities: Combine reference images, video, audio, documents, and web pages in one request, orchestrated by the prompt.
- Documents as a source: doc, xls, ppt, pdf, txt, key, pages, numbers, and md become teaching decks, product demos, animated charts, and business reports.
- Photorealistic world renderer: Faithful skin and facial detail with restrained, natural emotion — distinct faces rather than the same generic look, holding up even in crowd scenes.
- Pixel-level consistency: Characters, props, sound, and spatial relationships stay aligned to your references — not "close enough", but faithfully reproduced.
- Strict first/last frame mode: Pin the exact opening and closing frames when you need controlled transitions.
- Native audio output: Generated video carries its own audio track; the switch is free either way.
- Adaptive aspect ratio: Let the model pick the ratio from your intent and input media, or choose from five presets.
🛠️ Technical Specifications
| Item | Details |
|---|---|
| Model architecture | All-in-one omni-reference video generation model |
| Task type | Video generation (text / image / video / audio / document to video) |
| Prompt | Chinese and English, up to 5000 characters. Either prompt or reference media is required |
| Output Format | Video with native audio |
| Resolution | 480P, 720P, 1080P (default 1080P) |
| Duration | 2–30 seconds (default 5); -1 enables smart duration. With video input, input plus output must not exceed 30 seconds |
| Aspect ratio | adaptive (default), 16:9, 4:3, 1:1, 3:4, 9:16 |
| Reference images | Up to 10; JPEG, JPG, PNG (no alpha), BMP, WEBP; ≤20MB each; 240–8000px per side; ratio within 8:1 |
| Reference video | Up to 5 clips, 1–15s each, 15s total; mp4, mov; <100MB each; 240–4096px per side |
| Reference audio | Up to 5 clips, 1–15s each, 15s total; wav, mp3; <15MB each |
| Documents | One file, ≤100MB, ≤50 pages; docx, doc, xlsx, xls, pptx, ppt, pdf, txt, key, pages, numbers, md. Requires thinking mode |
| Web pages | One public URL, no login required. Requires thinking mode; mutually exclusive with documents |
| First / last frame | One image each; mutually exclusive with omni-reference inputs |
| Thinking mode | Off by default; required for documents, web pages, and complex image reasoning |
| Seed | Optional; 0–2147483647 for reproducibility |
| Audio switch | On by default; pricing is unchanged either way |
| Latency | No cold start |
Sample Prompts
Video 1 holds Image 3 and plays a gentle country ballad on the chair from Image 4, saying "the sunshine is lovely today". Image 1 walks past holding Image 2, sets it on the table beside Video 1, and says "that sounds great, could you play it again".A cinematic 30-second one-take through a rainy neon alley at night, camera drifting forward past shop signs and reflections, ambient city sound with distant traffic.Turn the attached product deck into an animated explainer: highlight each key metric as an on-screen chart, keep the brand colors, calm narrator pacing.
💰 Pricing
Billed by video seconds, counting both input and output video.
Billable duration = input video duration + output video duration
| Resolution | Price per Second |
|---|---|
| 480P | $0.0625 |
| 720P | $0.125 |
| 1080P | $0.25 |
Example Costs
| Scenario | Billable Duration | 480P | 720P | 1080P |
|---|---|---|---|---|
| 10s output, no reference video | 10s | $0.625 | $1.25 | $2.50 |
| 15s output + 5s reference video | 20s | $1.25 | $2.50 | $5.00 |
| 30s output, no reference video | 30s | $1.875 | $3.75 | $7.50 |
Billing Rules
- Input video and output video are both billed, by video seconds.
- Input video billable duration is the actual input seconds; output is the successfully generated seconds.
- Reference images and reference audio are not billed by duration.
- Turning the audio track on or off does not change the price.
- Failed tasks are not billed and are refunded in full.
- Smart duration (
-1) is pre-charged at the 30-second cap and settled to the actual generated length on completion.
💡 Best Use Cases
- Complete short-form storytelling: 30 seconds of continuous video with room for pacing, continuous camera moves, and one-take shots.
- Office material to video: Turn decks, spreadsheets, and PDFs into teaching content, product demos, animated charts, and business reports.
- Character and style consistency: Hold a character, wardrobe, prop, or voice steady across a series using multi-modal references.
- E-commerce and advertising: Build campaign visuals from product shots, brand key art, and reference footage in a single request.
- Controlled transitions: Use strict first/last frame mode when the opening and closing frames must be exact.
🔗 Related Models
- Wan 2.7 Image to Video: A lighter option when you only need to animate a single reference image.
- Wan 2.7 Text to Video: A better fit for short clips generated purely from a prompt.