WAN 2.7 video starts to stand out when a short clip needs stable motion, readable camera direction, and clean image to video animation from one still frame.
WAN 2.7 generates short videos from a text prompt, and it can also animate a still image into video. Below are 5 real runs (four text-to-video, one image-to-video) uploaded as MP4s so you can watch the results directly.
Cover art (generated)
Base frame source: Test 1. Prompt: A highly detailed macro shot of a crystal clear glass on a rustic wooden table. Thick steaming hot espresso is slowly poured into cold milk, creating physically accurate swirling gradients and foam. Warm golden hour light, shallow depth of field, cinematic.
Test setup
Video model
alibaba/wan-2-7
Duration
5 seconds (all tests)
Resolution
720P
Prompt Extend
false (kept prompts as-written)
Watermark
false
Observed cost
$0.50 per 5s run at 720P (from Wiro run cost)
5 video prompt tests
Test 1: Espresso macro physics (text-to-video)
Prompt: A highly detailed macro shot of a crystal clear glass on a rustic wooden table. Thick steaming hot espresso is slowly poured into cold milk, creating physically accurate swirling gradients and foam. Warm golden hour light, shallow depth of field, cinematic. Negative: blurry, low quality, artifacts.
Test 2: Human hands + motion (text-to-video)
Prompt: Cinematic handheld shot in a subway station. A street violinist plays under warm fluorescent lights. The camera slowly dolly-ins from waist-up to close-up of the hands on the strings. Natural crowd movement in the background, realistic motion blur. Negative: glitch, warped hands, artifacts.
Test 3: Macro tracking + environment motion (text-to-video)
Prompt: Macro cinematic shot of a tiny friendly robot walking through tall grass after rain. Dew drops sparkle. The camera tracks alongside at ground level. Natural wind moves blades of grass, shallow depth of field, soft bokeh. Negative: glitch, extra limbs, artifacts.
Test 4: Landscape + speed of motion (text-to-video)
Prompt: Wide aerial shot of a snowy mountain ridge at sunrise. Thin clouds drift quickly across the peaks like a gentle time-lapse. Warm light spills over the ridge, subtle lens flare, stable horizon, cinematic. Negative: glitch, jitter, artifacts.
Test 5: Animate a still image (image-to-video)
Input image:
Input image used for image-to-video.Prompt: The scene comes alive with a gentle breeze. The calm green water begins to ripple naturally, softly distorting the crisp reflections of the brightly colored buildings. The moored boats gently bob up and down against their wooden poles, creating small realistic wakes. The camera executes a very slow stable push in straight down the center of the canal.
Quick takeaways
For short 5-second clips, you can iterate quickly on camera language (dolly-in, tracking, aerial).
Negative prompts help keep runs away from obvious artifacts, but they do not guarantee perfect anatomy.
For image-to-video, the input image drives the output framing and ratio.
