MiniMax H3 视频生成

multipart：图片是 0.00 秒的 <Picture 1> 首帧。

curl -fsS https://minimax.api.easyframe.cn/v1/videos \
  -H "Authorization: Bearer $MINIMAX_API_KEY" \
  -F model=minimax-h3-base \
  -F mode=i2va \
  -F resolution=480p \
  -F seconds=5 \
  -F aspect_ratio=16:9 \
  --form-string prompt_optimization=false \
  --form-string prompt='For the target video, at 0.00 seconds into the target video, <Picture 1> (from [Shot 1]) is fully referenced.

integrated_multimodal_description: [Shot 1] The camera slowly pushes toward the woman in <Picture 1> as she looks up and smiles.

overall_soundscape: Soft room tone and fabric movement.

non_diegetic_music: N/A' \
  -F images=@first.png

multipart：第一张是首帧，第二张是可选尾帧。

curl -fsS https://minimax.api.easyframe.cn/v1/videos \
  -H "Authorization: Bearer $MINIMAX_API_KEY" \
  -F model=minimax-h3-base \
  -F mode=fl2va \
  -F resolution=480p \
  -F seconds=5 \
  -F aspect_ratio=16:9 \
  --form-string prompt_optimization=false \
  --form-string prompt='How the reference pictures align with the target video — Picture 1 (from Shot 1) aligns with the 0.00-second mark of the target video; Picture 2 (from Shot 1) aligns with the 5.00-second mark of the target video.

integrated_multimodal_description: [Shot 1] A product begins in the position and framing established by Picture 1. The camera holds a static shot as the product rotates at a steady speed and settles into the exact pose, spacing, and composition established by Picture 2 at the end.

overall_soundscape: A quiet room tone continues beneath the soft mechanical sound of the rotating platform.

non_diegetic_music: N/A' \
  -F images=@start.png \
  -F images=@end.png

multipart：图片是视频结尾的 <Picture 1> 尾帧。

curl -fsS https://minimax.api.easyframe.cn/v1/videos \
  -H "Authorization: Bearer $MINIMAX_API_KEY" \
  -F model=minimax-h3-base \
  -F mode=l2va \
  -F resolution=480p \
  -F seconds=5 \
  -F aspect_ratio=9:16 \
  --form-string prompt_optimization=false \
  --form-string prompt='How the reference pictures align with the target video — <Picture 1> (from [Shot 1]) aligns with the 5.00-second mark of the target video.

integrated_multimodal_description: [Shot 1] A falling glass breaks and its fragments settle into the exact composition in <Picture 1>.

overall_soundscape: A sharp glass impact followed by fragments sliding.

non_diegetic_music: N/A' \
  -F images=@last.png

multipart：参考图像和音频字段可重复，提示词使用六段式。

curl -fsS https://minimax.api.easyframe.cn/v1/videos \
  -H "Authorization: Bearer $MINIMAX_API_KEY" \
  -F model=minimax-h3-fast \
  -F mode=ref2va \
  -F resolution=720p \
  --form-string prompt_optimization=false \
  --form-string prompt='subject_definitions:
<Subject 1> is the person whose appearance comes from <Picture 1>.
<Audio 1> is the voice-timbre reference for <Subject 1> (S1).

summary:
[reference generation + audio reference] The target video preserves <Subject 1> and follows the referenced voice timbre.

retention_analysis:
<Subject 1> (appears in [Shot 1]): fully_preserved - identity and clothing are retained.
<Audio 1>: reference - voice timbre is referenced without copying the signal.

detailed_description:
The target video is in a realistic cinematic style.
[Shot 1] <Subject 1> (S1) walks through the frame and says, <d>[Chinese] 我们出发吧。</d>

overall_soundscape:
Natural outdoor ambience and footsteps.

non_diegetic_music:
N/A' \
  -F seconds=6 \
  -F aspect_ratio=16:9 \
  -F images=@person.png \
  -F audios=@voice.wav