豆包拍照读说明书:用API引用原文生成操作卡

apizt 实测记录 · 2026-10-24 · 不是第三方产品界面截图

通过apizt视觉模型实际读取英文手册和设备图。第一稿漏掉电源长按时长;加入逐条原文引用后,结果正确保留POWER长按2秒、FEED按一次、STATUS是指示灯与TEST 01测试标签。

本次操作

  1. 分开文档与设备图:上传本例虚构Q-LABEL01英文手册及示意图,先核对左右按钮位置。
  2. 要求引用原文和数字:让模型保留英文引句、中文动作、hold_seconds和状态含义,缺失项用null,不补WiFi等功能。
  3. 逐条与手册复核:核对先关机开盖、纸卷朝前、关盖、长按开机、绿灯、按FEED的顺序;第一稿的2秒遗漏保留作负例。

结果与限制

api-vision

httpmodelrequestSeconds
200gpt-4.1-mini6.696
查看原始模型输出
{
  "labels": [
    {
      "label": "POWER",
      "chinese": "电源",
      "position": "左侧第一个按钮",
      "source_quote": "POWER"
    },
    {
      "label": "FEED",
      "chinese": "进纸",
      "position": "中间按钮",
      "source_quote": "FEED"
    },
    {
      "label": "STATUS",
      "chinese": "状态指示灯",
      "position": "右侧小圆灯",
      "source_quote": "STATUS"
    },
    {
      "label": "TEST 01",
      "chinese": "测试标签",
      "position": "出纸口正中",
      "source_quote": "The test label reads TEST 01."
    }
  ],
  "steps": [
    "1. 关闭打印机电源并打开顶盖。",
    "2. 放置纸卷,使标签从前方出纸口方向出来。关闭顶盖。",
    "3. 打开电源。等待状态指示灯变为绿色。",
    "4. 状态指示灯为绿色时,按一次进纸按钮。",
    "5. 观察打印出的测试标签,内容为 TEST 01。"
  ],
  "unknown_features": [
    "未提供切纸功能说明。",
    "未提供蓝牙、Wi-Fi、手机应用、固件升级等功能说明。"
  ]
}

api-vision-timing

httpmodelrequestSeconds
200gpt-4.1-mini13.09

视觉API实际操作卡

namepositionoperation
POWER左侧第一个按钮按住2秒开机
FEED中间按钮按一次,走一张标签
STATUS右侧指示灯绿色表示准备好,仅指示灯
  1. 关闭打印机,打开顶盖。
  2. 放入纸卷,使标签从前方槽口出来,关闭顶盖。
  3. 开机,STATUS绿灯亮时,按FEED一次。测试标签显示“TEST 01”。
查看原始模型输出
{
  "source_english_quotes": {
    "purpose": "Q-LABEL 01 is a fictional teaching printer. This manual describes only loading a paper roll and printing a test label.",
    "buttons": {
      "POWER": "Hold for 2 seconds to turn on.",
      "FEED": "Press once to move one label.",
      "STATUS": "A green indicator means ready. It is an indicator, not a button."
    },
    "load_a_roll": [
      "Turn the printer off and open the top cover.",
      "Place the roll so labels exit toward the front slot. Close the cover.",
      "Turn on. When STATUS is green, press FEED once. The test label reads TEST 01."
    ],
    "not_specified": "No Wi-Fi setup, Bluetooth pairing, phone app, firmware update or cutting function is defined in this teaching model."
  },
  "chinese_steps": {
    "buttons": [
      {
        "name": "POWER",
        "position": "左侧第一个按钮",
        "operation": "按住2秒开机"
      },
      {
        "name": "FEED",
        "position": "中间按钮",
        "operation": "按一次,走一张标签"
      },
      {
        "name": "STATUS",
        "position": "右侧指示灯",
        "operation": "绿色表示准备好,仅指示灯"
      }
    ],
    "load_a_roll": [
      "关闭打印机,打开顶盖。",
      "放入纸卷,使标签从前方槽口出来,关闭顶盖。",
      "开机,STATUS绿灯亮时,按FEED一次。测试标签显示“TEST 01”。"
    ]
  },
  "buttons": {
    "POWER": {
      "hold_seconds": 2,
      "operation": "Hold for 2 seconds to turn on"
    },
    "FEED": {
      "press_count": 1,
      "operation": "Press once to move one label"
    },
    "STATUS": {
      "type": "indicator",
      "state_meaning": "green means ready"
    }
  },
  "indicator_states": {
    "STATUS": {
      "green": "ready"
    }
  },
  "loading_sequence": [
    "Turn the printer off",
    "Open the top cover",
    "Place the roll so labels exit toward the front slot",
    "Close the cover",
    "Turn on",
    "When STATUS is green, press FEED once",
    "Test label reads TEST 01"
  ],
  "test_print_text": "TEST 01",
  "unsupported_functions": [
    "Wi-Fi setup",
    "Bluetooth pairing",
    "phone app",
    "firmware update",
    "cutting function"
  ]
}

复跑与凭据

下载本题 evidence.zip,保留各任务目录。API脚本通过本机环境变量 tokenba_API_KEY 读取你的Key;不把Key放进请求JSON。先核对模型广场当前可用型号和价格,再显式运行一次小样。

生成次数、文件成功、内容质检分别判断。没有在公开报告中保存任何可用Key。