git clone https://github.com/Comfy-Org/ComfyUI.git
cd ComfyUI
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python main.py

export MINIMAX_H3_WORKDIR="$PWD/minimax-h3-deploy"
mkdir -p "$MINIMAX_H3_WORKDIR"

git clone https://github.com/sgl-project/sglang.git "$MINIMAX_H3_WORKDIR/sglang"
cd "$MINIMAX_H3_WORKDIR/sglang"
git checkout 2511743bd784e69e5a81ca3d926a000711dae4ab

python3 -m pip install --upgrade uv
uv venv --python 3.12 .venv
source .venv/bin/activate
uv pip install -e "python[diffusion]" --prerelease=allow
uv pip install "huggingface_hub==1.28.0"
sglang --version

export HF_HOME="$MINIMAX_H3_WORKDIR/hf-cache"
export MINIMAX_H3_MODEL_DIR="$MINIMAX_H3_WORKDIR/model"

hf download MiniMaxAI/MiniMax-H3 \
  --revision 42ed227ee7df40d41602854ae760620d6eb651fe \
  --include "model_index.json" "FL2VA/*" \
  --local-dir "$MINIMAX_H3_MODEL_DIR"

HF_HUB_OFFLINE=1 sglang serve \
  --model-path "$MINIMAX_H3_MODEL_DIR" \
  --model-variant fl2va \
  --num-gpus 8 \
  --tp-size 1 \
  --ulysses-degree 8 \
  --encoder-parallel auto \
  --performance-mode speed \
  --host 127.0.0.1 \
  --port 30010

curl --fail --retry 60 --retry-delay 10 --retry-all-errors \
  http://127.0.0.1:30010/health

video_id=$(
  curl -fsS -X POST http://127.0.0.1:30010/v1/videos \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMaxAI/MiniMax-H3",
    "prompt": "At night, three cats perform with tiny brass instruments in a warm living room, with movement synchronized to the music.",
    "seconds": 5,
    "task": "t2va",
    "conditions": [],
    "target": {
      "short_edge": 768,
      "aspect_ratio": "16:9",
      "duration_seconds": 5.0
    },
    "num_outputs_per_prompt": 1,
    "num_inference_steps": 50,
    "flow_shift": 12.0,
    "audio_flow_shift": 3.0,
    "seed": 1101
  }' | jq -r '.id'
)
test -n "$video_id" && test "$video_id" != "null"

while true; do
  job=$(curl -fsS "http://127.0.0.1:30010/v1/videos/${video_id}")
  status=$(printf '%s' "$job" | jq -r '.status')
  case "$status" in
    completed) break ;;
    failed)
      printf '%s\n' "$job" | jq . >&2
      exit 1 ;;
    *) sleep 5 ;;
  esac
done

curl -fsS -L "http://127.0.0.1:30010/v1/videos/${video_id}/content" -o minimax-h3-t2va.mp4

ffprobe -v error \
  -show_entries stream=codec_type,codec_name,width,height,r_frame_rate,sample_rate,channels \
  -of json minimax-h3-t2va.mp4

# 下载 Ref2VA 分区
hf download MiniMaxAI/MiniMax-H3 \
  --revision 42ed227ee7df40d41602854ae760620d6eb651fe \
  --include "model_index.json" "Ref2VA/*" \
  --local-dir "$MINIMAX_H3_MODEL_DIR"

# 启动 Ref2VA 服务（另一端口）
HF_HUB_OFFLINE=1 sglang serve \
  --model-path "$MINIMAX_H3_MODEL_DIR" \
  --model-variant ref2va \
  --num-gpus 4 \
  --ulysses-degree 4 \
  --performance-mode speed \
  --host 0.0.0.0 \
  --port 30011

pip install "torch>=2.4.0" \
  "diffusers>=0.32.2" \
  "transformers>=4.45.0" \
  "safetensors>=0.4.3" \
  "accelerate>=0.34.0" \
  "huggingface_hub>=0.25.0" \
  "numpy>=1.24.0" \
  "Pillow>=10.0.0" \
  "soundfile>=0.12.0" \
  "imageio>=2.34.0" \
  "imageio-ffmpeg>=0.5.0" \
  "av>=11.0.0"

import torch
from diffusers import ModularPipeline

pipe = ModularPipeline.from_pretrained(
    "MiniMaxAI/MiniMax-H3",
    torch_dtype=torch.bfloat16,
    trust_remote_code=True,
)
pipe.to("cuda")

result = pipe(
    prompt="A cinematic shot of a robot walking through a neon-lit city at night, rain reflecting on the street.",
    negative_prompt="",
    width=1344,
    height=768,
    num_inference_steps=50,
    guidance_scale=1.0,
    num_frames=124,          # 约 5 秒 @ 24FPS
    seed=42,
    audio=True,              # 同时生成立体声音频
)

result.videos[0].save("output.mp4")
result.audios[0].save("output_audio.wav")

curl -fsS -X POST http://localhost:30010/v1/videos \
  -H "Content-Type: application/json" \
  -d '{
    "task": "t2va",
    "prompt": "Cinematic, medium wide shot. A lone astronaut walks across a vast alien desert under two moons, dust kicking up behind her boots, a low synth score building.",
    "conditions": [],
    "target": {
      "short_edge": 768,
      "aspect_ratio": "16:9",
      "duration_seconds": 10
    },
    "seed": 0
  }'

curl -fsS -X POST http://localhost:30010/v1/videos \
  -H "Content-Type: application/json" \
  -d '{
    "task": "i2va",
    "prompt": "Continue the scene naturally: the woman turns toward the window and smiles.",
    "conditions": [
      {
        "media": "<BASE64_ENCODED_FIRST_FRAME>",
        "media_type": "image"
      }
    ],
    "target": {
      "short_edge": 768,
      "aspect_ratio": "16:9",
      "duration_seconds": 5
    },
    "seed": 0
  }'

curl -fsS -X POST http://localhost:30010/v1/videos \
  -H "Content-Type: application/json" \
  -d '{
    "task": "ref2va",
    "prompt": "<Picture 1> provides the character identity and outfit style. <Video 1> provides the camera motion. Continue the story: the hero draws his sword and faces the dragon.",
    "conditions": [
      { "media": "<BASE64_PICTURE>", "media_type": "image" },
      { "media": "<BASE64_VIDEO>", "media_type": "video" }
    ],
    "target": {
      "short_edge": 768,
      "aspect_ratio": "16:9",
      "duration_seconds": 5
    },
    "seed": 0
  }'

integrated_multimodal_description: [Shot 1] Cinematic, ...（分镜描述：景别、运镜、动作、时间点）
overall_soundscape: A low, resonant hum ...（整体声景：环境声、对白、音效）
non_diegetic_music: Cinematic space-opera orchestral score ...（非叙事配乐：曲风、节奏、情绪）

# SGLang 服务地址
SGLANG_DEPLOYMENT_URL="<sglang-deployment-url>"

# MiniMax API（二选一）
# 国内
MINIMAX_API_BASE="https://api.minimaxi.com"
# 国际
# MINIMAX_API_BASE="https://api.minimax.io"

# 平台创建 API key
TOKEN="<token>"