Creating short video ads for Taiwanese brands with Veo 3 (Sora on the same endpoint coming soon) — complete 5-minute tutorial

For Taiwanese brands making Reels / TikTok / IG short video ads, the biggest pain point is usually that finding a crew to film, edit, and post-produce takes a week. AI text-to-video compresses this workflow into 5 minutes, and the quality is now good enough for real commercial campaigns. This article shows you how with practical examples.

The live text-to-video model currently available on Kunavo is Google Veo 3; OpenAI’s Sora (sora-2) is still on the roadmap and is not yet available. The good news is that both use the same OpenAI-compatible endpoint, so all examples in this article use veo-3; once Sora launches, you’ll only need to change one model field. See the full details in the Sora API guide.

Create a 9:16 vertical short video in 5 minutes

Kunavo wraps Veo 3 in an OpenAI-compatible /v1/videos endpoint, so you can use it by changing one line: base_url

import time
from openai import OpenAI

client = OpenAI(
    api_key="sk-kn-...",
    base_url="https://api.kunavo.com/v1",
)

def make_video(**params):
    """送出影片任務，等它完成，回傳影片 URL。"""
    # OpenAI SDK 沒有能直接呼叫這個端點的方法，所以用同一個 client 送 JSON
    task = client.post("/videos", body=params, cast_to=object)
    while task["status"] not in ("completed", "failed"):
        time.sleep(5)  # 生成通常要 30 秒到幾分鐘
        task = client.get(f"/videos/{task['id']}", cast_to=object)
    if task["status"] == "failed":
        raise RuntimeError(task["error"]["message"])
    return task["output"]["url"]

# 文字轉影片：8 秒、9:16 直式（Reels / TikTok / IG 短影片）
# 每支 Veo 影片固定 8 秒，Veo 模型會忽略 duration
url = make_video(
    model="veo-3",
    prompt=(
        "近距離鏡頭：手沖咖啡從濾杯滴入透明玻璃杯，"
        "蒸氣緩緩上升，深色背景，柔光打在杯緣上。"
        "結尾 logo 浮現：『晨曦咖啡 · 台北東區』"
    ),
    aspect_ratio="9:16",
    resolution="1080p",
)
print(url)  # files.kunavo.com 的永久連結

Every Veo 3 video is fixed at 8 seconds, and a 1080p vertical video costs $0.42 (Veo 3 Quality on Kunavo is $1.60 and Veo 3 Lite $0.224). That’s two orders of magnitude cheaper than hiring a local Taiwanese production crew.

Image-to-video — bring existing product images to life

Your momo / Shopee / independent store already has product photos. Veo 3 supports using an image as the first frame, and AI automatically generates an 8-second orbiting shot or detail showcase:

# 圖生影片：用既有產品圖做 8 秒展示影片（沿用上面的 make_video）
url = make_video(
    model="veo-3",
    prompt="鏡頭環繞商品旋轉，柔和光線變化",
    image_url="https://your-cdn.com/product-hero.jpg",
    aspect_ratio="9:16",
)
print(url)

A product photographer charges about NT$500-1500 for one product image, plus at least NT$3000-5000 for a moving shot; Veo 3 costs $0.50, about NT$16 at the exchange rate, so the whole video costs just NT$8.

5 common applications for Taiwanese brands

1. New drink launch: Hand-pouring, a milk foam flip, syrup drizzle — describe it in text to generate multiple 8-second video versions and A/B test click-through rates on FB Ads
2. Night market food: Sizzling oil, a close-up as food comes out of the fryer, toppings being poured on — combine with local Taiwanese scenes (red night market lights, old street signs) to create brand recognition
3. E-commerce product Reels: Product photo → Veo 3 image-to-video → post directly to IG / TikTok / Shopee product pages
4. Tourism / hospitality: Hotel rooms, aerial views of attractions, seasonal atmosphere — no need to send a photographer to Alishan; AI generation is good enough
5. Unboxing video intro for individual creators: A high-quality 8-second intro, then use Suno on Kunavo to generate an original theme song for the background music

A few things to keep in mind

• Traditional Chinese text rendering: Veo 3 is currently still unreliable at rendering Traditional Chinese characters in video. We recommend adding subtitles during post-production (CapCut / the Taiwan version of Jianying) rather than having AI write text
• Real human faces: All top models have restrictions on “real celebrities / political figures.” For commercial use, focus on products, scenes, and cartoon characters to avoid infringement disputes
• Music: Veo 3 generates ambient audio itself. For professional BGM, pair it with original music generated by Suno on Kunavo
• Video links do not expire: Kunavo stores every video on its own CDN, and the files.kunavo.com URL it returns is permanent. For a product in production, it is still worth keeping a copy in your own S3 / Bunny CDN / Cloudinary

Payment — Taiwanese bank cards accepted

Visa / Mastercard / JCB / American Express are all supported. Minimum top-up: $10 USD, automatically converted to TWD at Stripe checkout. No subscription, pay by usage, and your balance never expires. A $10 top-up is enough to generate 20 Veo 3 videos or about 4 Veo 3 Quality videos.

Advanced: how to reliably produce commercially usable quality

• Generate in batches + select: Run the same prompt 3-5 times and pick the best result. AI video is random; the chance of getting it right on the first try is about 30%
• Detailed prompts beat simple prompts: “coffee cup” is poor; “light wood tabletop, backlighting, steam, slow motion, advertising style” is good
• Use negative words to remove common flaws: “avoid distorted faces, avoid mangled text, no shaky camera” (negative words need to be included in the prompt)

Ready to get started? Sign up for free, top up from $10, and pay by usage. See the full video API documentation at /docs/video and browse all video models at /models.