Wan (万相) is Alibaba's text-to-video model family — Wan 2.2, 2.5, and the current Wan 2.7 — known for strong cinematic composition and motion. The Wan API generates short clips from a prompt programmatically, and Wan 2.7 is live on Kunavo through the OpenAI-compatible /v1/video/generations endpoint: one bearer token, transparent per-second pricing, no Alibaba Cloud account.
Wan API pricing
Wan 2.7 on Kunavo bills per second of generated video, by resolution:
| Resolution | Kunavo | Alibaba list | 5s clip |
|---|---|---|---|
| 720p | $0.096 / second | $0.10 / second | $0.48 |
| 1080p | $0.144 / second | $0.15 / second | $0.72 |
Honest note: the per-second rate is only slightly under Alibaba's own list — if Wan is the only model you need, going direct is a fine choice. The reason to run it on Kunavo is operational: the same key and balance also drive Veo 3, image models, Suno and the Claude / GPT / Gemini text lineup, with per-clip USD line items instead of provider credits.
Generate a video with Wan 2.7
Submit a prompt to POST /v1/video/generations with model: "wan-2-7". Generation takes a few minutes; in production use the async /v1/videos task lifecycle with idempotency keys.
import requests
resp = requests.post(
"https://api.kunavo.com/v1/video/generations",
headers={"Authorization": f"Bearer {API_KEY}"},
json={
"model": "wan-2-7",
"prompt": "a red lantern swaying in a narrow Kyoto alley at dusk, rain",
"duration": 5,
"aspect_ratio": "16:9",
"resolution": "720p", # or "1080p"
},
timeout=600,
)
print(resp.json()["data"][0]["url"])Wan 2.2 API and Wan 2.5 API
Searches for the Wan 2.2 API and Wan 2.5 API refer to earlier generations of the same family — 2.2 introduced the MoE architecture and the Animate variants, 2.5 added native audio experiments. Wan 2.7 supersedes both on quality and instruction following. The request shape is identical across generations, so a Wan 2.2/2.5 integration ports to Kunavo with a one-word change to the model field.
Wan vs Veo 3
Both are live on the same endpoint, so the honest answer is: A/B test your prompts. Rules of thumb — Wan 2.7 is the budget cinematic pick with per-second billing ($0.48 per 5s 720p clip); Veo 3 leads overall quality and generates native synchronized audio, from $0.16 (Lite) to $1.92 (Quality) per 8s 720p clip. For the whole landscape — Veo, Sora, Kling, Hailuo, Seedance — see the text-to-video API guide.
FAQ
What is the Wan API?
Alibaba's text-to-video family, live on Kunavo as Wan 2.7 via the OpenAI-compatible /v1/video/generations endpoint.
How much does it cost?
$0.096/second at 720p and $0.144/second at 1080p — a 5-second 720p clip is $0.48, billed per clip in USD.
Do I need an Alibaba Cloud account?
No — one Kunavo key (sk-kunavo-…) and a Stripe balance; the same key reaches Veo 3, image, music and text models.
Does Wan 2.7 do image-to-video?
Image-to-video and the Wan image-edit line (Wan 2.7 Image Pro) are on the roadmap; text-to-video is what's live today. Pass image_url workflows through Veo 3 in the meantime.