MiniMax H3 (Hailuo-03) API

规格 / 通道	官方	云厂商
2K	$0.13	$0.091
768p	$0.08	$0.056

规格 / 通道	官方	云厂商
2K	$0.13	$0.091
768p	$0.08	$0.056

规格 / 通道	官方	云厂商
2K	$0.13	$0.091
768p	$0.08	$0.056

Input

first_frame_url

Drag and drop or click to uploadJPEG, PNG, WEBP · 30 MB

File uploads require configured image storage. A public HTTPS image URL can be used directly.

last_frame_url

Drag and drop or click to uploadJPEG, PNG, WEBP · 30 MB

File uploads require configured image storage. A public HTTPS image URL can be used directly.

prompt *

Use Image 1 as the exact opening frame and Image 2 as the exact ending frame. Create a 12-second architectural brand reveal that transforms the closed pavilion into the fully opened MiniMax H3 API and Kie.ai exhibition.
[0–3 seconds]
Hold the opening composition almost still. Let reflections shift subtly across the frosted glass while the exterior architectural lights turn on section by section. Keep “MiniMax H3 API” and “Kie.ai” readable and unchanged.
[3–7 seconds]
The central doors begin opening slowly and symmetrically. Warm interior light appears through the widening gap and travels across the plaza. Introduce gentle camera movement by pushing forward only slightly; do not change the original frontal alignment.
[7–10 seconds]
As the doors open wider, the interior fabric installations move softly from indoor airflow. Sculptural panels slide into alignment and reveal portions of the final sentence in a controlled sequence. Do not replace the architecture or invent a different interior.
[10–12 seconds]
Arrive precisely at Image 2. The full line “MiniMax H3 API is now available on Kie.ai” becomes completely visible and stable. Hold the ending frame with only subtle fabric movement and light reflections.
Audio: quiet city ambience, low architectural room tone, soft mechanical door movement, subtle fabric sounds, and one refined musical swell as the final line is revealed.
Preserve the building geometry, camera position, text spelling, plaza layout, lighting direction, and all core visual elements. No people, hard cuts, rapid zooms, morphing architecture, floating UI, extra text, misspellings, glowing cyber effects, or camera rotation.

Video description prompt1662/7000

duration *

Video duration in seconds. Available values: 4-15.

resolution

Video resolution.