curl-X POST 'https://pollo.ai/api/platform/v1/generation/alibaba/wan-v3-0/video'\-H'Content-Type: application/json'\-H'x-api-key: YOUR_API_KEY'\-d'{ "input": { "prompt": "A cinematic shot of a golden retriever running through a field of sunflowers at sunset, warm rim light, shallow depth of field", "duration": 5, "resolution": "480p", "aspectRatio": "adaptive", "generateAudio": true }}'
{"input":{"prompt":"A cinematic shot of a golden retriever running through a field of sunflowers at sunset, warm rim light, shallow depth of field","duration":5,"resolution":"480p","aspectRatio":"adaptive","generateAudio":true},"webhookUrl":"https://example.com/webhooks/pollo"}
模式
輸入所選端點和模式的欄位。
場地
類型
必需的
描述
prompt
string
是的
Text prompt describing the content to generate.
duration
integer
不
The duration of the generated video, in seconds.
resolution
string
不
Output resolution of the generated media (e.g. 720p, 1080p, 2K, 4K). Values are case-insensitive.
aspectRatio
string
不
Output aspect ratio, as width:height (e.g. 16:9).
generateAudio
boolean
不
Generate natural-sounding, fitting audio for the output video.
Recurring Character Stories: Pair a character reference with a new setting and a short sequence of actions for the next installment.
Location-Based Narratives: Use a location image to anchor a scene that develops from an establishing view into a character interaction.
Audio-Led Scene Concepts: Explore footage built around a supplied sound reference, with the relationship between sound and action specified in the brief.
Alternative Scene Endings: Develop different reactions to the same event while keeping the preceding story context consistent.
Craft Process Stories: Show a maker inspecting a material, working on it, and revealing a result within a connected narrative.
Reference-Based Previsualization: Turn approved visual references into motion studies before committing resources to a live shoot.
如何使用 Wan 3.0 API
Get an API Key: Create a Pollo API account and generate your API key from the developer dashboard.
Choose the Model: Send requests to the Wan 3.0 endpoint at /generation/alibaba/wan-v3-0/video
Add Your Inputs: Provide your prompt and supported media. Set duration, resolution, and generateAudio as needed; follow the selected mode’s current schema.
Generate and Retrieve: Submit the request, retain the returned taskId, and check task status to retrieve the completed video.
Wan 3.0 API 提示詞最佳實踐
將場景寫成一個帶有明顯變化的小故事。為每個參考分配一個明確的目的，然後解釋連接開頭和結尾的動作。
簡單的提示詞公式
Reference roles + setting + subject + opening action + turning point + ending reaction + camera direction + sound
你應該注意什麼
Reference Roles: Explain whether each supplied asset establishes a character, setting, movement, or sound; avoid asking one asset to fulfill conflicting roles.
Story Progression: Replace a list of unrelated events with one cause-and-effect sequence that viewers can follow.
Character Anchors: Repeat the few details that must remain recognizable, such as a red scarf or a distinctive shoulder bag.
Sound Priorities: Identify the important audible event and leave quieter background sound subordinate to it.
Revision Scope: When editing, distinguish the requested change from the visual elements that should remain fixed.
