Wan 3.0 API费用对比：ModelGo vs OpenRouter四组实测

选择API平台前，应同时比较输出效果和账单。本次四组生产需求在ModelGo的费用为$3.636364，在OpenRouter的费用为$3.400000；下文展示实际输出和费用。

Wan 3.0是什么？

本次通过视频API测试Wan 3.0的文生视频和首帧驱动能力。它适合空间揭示、施工可视化、车辆功能展示和游戏环境等重视主体连续性的场景。

测试方法与实际费用

本次测试包含四组独立需求：两组纯文字生成、两组参考图生成；在双平台可访问时，使用相同Prompt、参考文件和请求设置。ModelGo费用来自逐条匹配的已结算账单，OpenRouter费用来自已完成响应中的用量费用；表格展示模型输出费用，保留六位小数，不包含单独生成参考图的费用。

费用口径：本文比较模型调用费用，不是为账户充值所需的实际支出；不包含充值或支付手续费，包括另一篇价格对比文章计入的5.5%充值费。本文费用差额及比例仅按调用费用计算。

四个真实生产案例

为准确复现真实调用，本节保留当时提交的英文Prompt原文；中文说明用于解释生产目标。

模块化桥梁装配演示

应用行业：建筑施工/B2B可视化。

用模块化桥梁装配镜头展示施工可视化，检查结构连接、部件轨迹及最终空间关系；这是概念展示，不能代替工程验算。

实际提交的英文Prompt

Create a 10-second architectural visualization of a small prefabricated pedestrian bridge being assembled across a narrow landscaped canal. Use a fixed elevated three-quarter camera with coherent engineering scale. Seconds 0–3: two finished concrete abutments and a clear central gap are visible. Seconds 3–7: a single steel bridge module is lowered slowly by a crane from above until both ends rest on the abutments; the module must remain rigid and correctly aligned. Seconds 7–10: the lifting slings slacken and the camera holds the completed bridge, with two adult workers standing safely on the bank for scale. Use calm overcast daylight. No road traffic, collapsing structures, jumping construction stages, text or safety-certification claims. Deliver one coherent ten-second clip. Avoid speed ramps, unrelated cutaways and abrupt changes in camera direction; preserve the specified subject throughout.

ModelGo费用高6.95%。

电动SUV灯光展示

应用行业：汽车/功能广告。

用电动SUV灯光揭示短片表现车辆特征，重点检查车体结构、灯光变化和镜头运动的稳定性。

实际提交的英文Prompt

Create a 10-second premium automotive lighting film for an original dark-emerald electric SUV in a minimal concrete studio. Begin with the vehicle stationary in shadow and the camera low at the front-left corner. Seconds 0–3: one continuous front light bar illuminates smoothly from center to edges. Seconds 3–7: the camera moves laterally past the front wheel toward the driver’s door, revealing realistic metallic paint and a stable body silhouette. Seconds 7–10: the cabin ambient light comes on behind the closed window as the camera settles into a side profile. Keep the wheels stationary, panel gaps consistent and reflections coherent. No manufacturer logos, license plates, driving footage, text or performance claims. Deliver one coherent ten-second clip. Avoid speed ramps, unrelated cutaways and abrupt changes in camera direction; preserve the specified subject throughout.

ModelGo费用高6.95%。

通勤背包雨天到办公室转场

应用行业：箱包/零售广告。

保留背包的蓝灰织物、斜拉链、卷口和橙色挂环，先在雨中站台滑镜展示水滴，再叠化为办公室，最后由成人手提起；主体位置、形状与尺度须稳定。

这张原创参考图用于ModelGo；对OpenRouter可访问的模型，对照请求也使用同一文件。

实际提交的英文Prompt

Use the supplied original modular commuter backpack as the exact product reference. Preserve its slate-blue fabric, single diagonal front zipper, roll-top closure and orange side loop. Create a 10-second continuous shot beginning with the backpack on a bench under a transit shelter during light rain. Seconds 0–4: a slow camera slide reveals a few droplets on the fabric. Seconds 4–8: the background transitions through a clean match dissolve to a warm modern office, while the backpack stays in the same screen position, shape and scale. Seconds 8–10: a natural adult hand lifts it by the top handle. No additional pockets, changing zippers, invented waterproof ratings, logos or text. Keep the visual transition controlled and useful for a retail campaign. Deliver one coherent ten-second clip. Avoid speed ramps, unrelated cutaways and abrupt changes in camera direction; preserve the specified subject throughout.

ModelGo费用高6.95%。

地牢传送门游戏预告

应用行业：游戏/发布预告。

保留地牢门洞的六块拱石、台阶、壁灯和中央封印，缓慢推近后让封印旋转四分之一圈、双门开启，最后在紫色门光前停住；建筑保持刚性。

这张原创参考图用于ModelGo；对OpenRouter可访问的模型，对照请求也使用同一文件。

实际提交的英文Prompt

Animate the supplied original stone dungeon doorway into a 10-second game teaser. Preserve its six carved arch stones, shallow steps, two bronze sconces and central circular seal. Seconds 0–4: the camera slowly approaches the doorway as dust moves through narrow torchlight. Seconds 4–7: the circular seal rotates one quarter turn and the two stone door panels slide apart, revealing a restrained violet portal inside the same opening. Seconds 7–10: the portal light illuminates the existing carvings and the camera stops before crossing the threshold. Keep architecture rigid, lighting coherent and particle density low. No monsters, violence, licensed symbols, text or new architectural wings appearing. Prioritize a readable environmental reveal over frantic motion. Deliver one coherent ten-second clip. Avoid speed ramps, unrelated cutaways and abrupt changes in camera direction; preserve the specified subject throughout.

ModelGo费用高6.95%。

如何计算费用差额

对于实际交付规格一致的输出，费用降低比例的计算方式为(OpenRouter 实际费用 − ModelGo 已结算费用) / OpenRouter 实际费用 × 100。判断两次输出是否可比时，应检查实际返回的尺寸、时长和音轨。

开始使用ModelGo API

先安装requests并设置环境变量MODELGO_API_KEY。以下代码使用首个真实案例的端点、模型ID和参数，已在代码中直接写入实际Prompt。此示例提交异步视频任务，后续轮询及下载方式见代码下方。

import os, requests

prompt = 'Create a 10-second architectural visualization of a small prefabricated pedestrian bridge being assembled across a narrow landscaped canal. Use a fixed elevated three-quarter camera with coherent engineering scale. Seconds 0–3: two finished concrete abutments and a clear central gap are visible. Seconds 3–7: a single steel bridge module is lowered slowly by a crane from above until both ends rest on the abutments; the module must remain rigid and correctly aligned. Seconds 7–10: the lifting slings slacken and the camera holds the completed bridge, with two adult workers standing safely on the bank for scale. Use calm overcast daylight. No road traffic, collapsing structures, jumping construction stages, text or safety-certification claims. Deliver one coherent ten-second clip. Avoid speed ramps, unrelated cutaways and abrupt changes in camera direction; preserve the specified subject throughout.'
headers = {"Authorization": "Bearer " + os.environ["MODELGO_API_KEY"]}
headers["X-DashScope-Async"] = "enable"
payload = {'model': 'wan3.0-video',
 'input': {'prompt': prompt},
 'parameters': {'resolution': '720P', 'duration': 10, 'ratio': '16:9', 'audio': False}}
response = requests.post('https://api.modelgo.ai/api/v1/services/aigc/video-generation/video-synthesis', headers=headers, json=payload, timeout=420)
response.raise_for_status()
result = response.json()
print(result)  # Save the returned task ID for polling.

提交后保存响应中的output.task_id，通过GET https://api.modelgo.ai/api/v1/tasks/{task_id}轮询至成功或失败状态；成功后使用响应中的output.video_url下载视频。

音频请求与实际交付：原始记录中，两个平台均将音频开关设为false，但返回文件仍检测到音频流。“含音轨”仅说明文件中存在音频流，不代表已确认有可听声音或音频开关生效。代码保留真实请求参数；本次测试无法确定返回文件未体现该开关的原因。

该端点的首帧请求使用input.media，条目中设置type: "first_frame"，并在url中提供参考图。

ModelGo与OpenRouter的模型ID和媒体请求格式存在差异。可以复用创作需求，接入时请根据各平台文档调整端点、参考图结构、任务轮询和下载处理。

ModelGo API文档

在ModelGo使用Wan 3.0

使用本页介绍的Wan 3.0模型，请前往ModelGo的Models页面。

前往ModelGo使用Wan 3.0

参考资料

以下资料用于核对模型能力、接口与用量记录；案例需求由本次测试独立编写。

• 模型官方文档或系列展示：用于了解能力与创意方向；相关系列资料不代表其他变体的规格保证。
• OpenRouter API参考：用于核对请求约定及平台差异。
• OpenRouter用量与费用记录：用于理解用量信息；本文费用来自实际测试记录。
• OpenRouter模型目录API：在新接入前核对模型ID与目录属性。
• OpenRouter视频生成指南：用于了解媒体请求结构和工作流。