Auto Midjourney
v8Automate Midjourney Alpha 网页 image generation from Claude using the 认证d https://alpha.midjourney.com 会话. Use this 技能 whenever the user wants to 创建 Midjourney images, submit MJ prompts, default to Midjourney v8, 优化 Midjourney prompt syntax, or poll/下载 结果s from the Midjourney 网页 应用 instead of Discord. Also use it when the user 分享s Alpha 网页 请求 samples, cookies, channel IDs, or asks to reverse-engineer the Midjourney 网页site 工作流.
运行时依赖
安装命令
点击复制技能文档
Auto Midjourney
Use the user's own Midjourney Alpha 网页 会话 to submit imagine jobs and optionally poll for 结果s.
This 技能 is intended for conservative, user-triggered assistance rather than unattended bulk 自动化.
What this 技能 does Submits prompts to https://alpha.midjourney.com/API/submit-jobs Defaults to Midjourney v8 for new prompts unless the user explicitly 请求s another version Keeps 凭证s in 环境 variables instead of hardcoding them into the 技能 Supports a one-command flow through scripts/运行_imagine.py Can read https://alpha.midjourney.com/API/user-mutable-状态 to inspect current 网页 设置tings Can infer user_id and singleplayer_ from the 认证d cookie 应用lies local conservative throttling to reduce accidental 请求 bursts Includes scripts/mj_doctor.py for 设置up 验证 Includes an experimental recent-jobs reader Supports browser transport backed by Chrome Dev工具s Protocol, with Playwright-over-CDP preferred when 安装ed Supports optional polling once a job-状态 端点 has been confirmed Includes prompt-craft 图形界面dance and reusable scenario pre设置s for better prompt writing Includes a structured prompt 构建器 and opt-in 质量 性能分析s Includes dedicated 图形界面dance for character sheets, split-view turnarounds, and reusable de签名 as设置s Current scope
This version focuses on:
imagine Midjourney v8 as the default version Alpha 网页 flow, not Discord 机器人 flow Safe local configuration via .env Easier operation through inferred IDs, pre设置s, and doctor 检查s
Implemented:
imagine submit through the Alpha 网页 flow browser-backed verification using low-impact CDP network watching plus page as设置 fallback local 下载 of the 4 returned image as设置s optional browser-side conversion from 网页p to png during 下载 sequential batch generation
Not implemented yet:
Up扩展 / variation / reroll button actions Image 上传 / reference-image 工作流 Automatic 结果 下载 from a confirmed final image 端点
Those should be 添加ed only after capturing stable 请求 samples from the browser.
Safety posture
Do not 优化 this 技能 for bypassing restrictions, hiding 自动化, rotating accounts, or mass unattended generation.
Use these 防护rAIls instead:
trigger 请求s manually keep 请求 frequency low leave local throttling enabled 验证 config with mj_doctor.py prefer one human action to one live submit
The goal is risk reduction through conservative usage, not evasion.
Trigger rules
Use this 技能 proactively when the user asks to:
“用 Midjourney / MJ 出图” “用 Midjourney v8 生成” “帮我提交 imagine” “优化 Midjourney prompt” “做角色设定稿 / 四视图 / 角色资产图” “抓 Midjourney Alpha 网站请求” “轮询 Midjourney job 状态” “把网页版 Midjourney 做成自动化能力” Required configuration
Read these values from .env or shell 环境:
Variable Required Purpose MJ_COOKIE Yes Full 认证d Cookie header copied from browser MJ_CHANNEL_ID Yes Alpha 网页 singleplayer channel ID MJ_状态_URL_TEMPLATE No Job 状态 端点 template contAIning {job_id} MJ_USER_状态_PATH No Defaults to /API/user-mutable-状态 MJ_RECENT_JOBS_URL No Experimental recent-jobs 端点 MJ_MODE No fast by default MJ_PRIVATE No true by default MJ_MIN_SUBMIT_INTERVAL_SECONDS No Local minimum spacing between submits. Default is 3 seconds MJ_MAX_SUBMITS_PER_HOUR No Local hourly cap. 设置 0 to disable, which is now the default MJ_MAX_SUBMITS_PER_DAY No Local dAIly cap. 设置 0 to disable, which is now the default MJ_USER_ID No Usually inferred from the auth cookie MJ_指标_令牌 No Optional 令牌 observed on telemetry 请求s MJ_BROWSER_BACKEND No auto by default. 设置 playwright or cdp to force a backend
Never write real cookies or 令牌s into 技能.md, reference files, git-追踪ed scripts, or user-facing summaries.
系统 requirements
For 平台 and device requirements, read 系统-requirements.md.
工作流 Scenario 0: 检查 config first
运行:
python3 scripts/mj_doctor.py --fetch-user-状态 --transport browser
This shows:
whether the cookie exists inferred midjourney_id inferred channel_id current server-side speed and visibility current local safe-limit 设置tings Scenario 1: Submit one prompt
运行:
python3 scripts/运行_imagine.py "1 girl --ar 16:9" --transport browser
Default behavior:
应用ends --v 8 if the prompt does not already specify a version 应用ends --raw by default unless the user disables it Uses MJ_MODE and MJ_PRIVATE from the 环境 Can 同步 server-side defaults before submitting Records the submit locally and enforces conservative pacing Prints structured JSON with the 请求 payload, submission 响应, and 提取ed job_id
For simplest live use:
python3 scripts/运行_imagine.py "cinematic portrAIt of a fox astronaut" --transport browser --同步-user-状态 --wAIt-page-as设置s --下载 --convert-to png
When --wAIt-page-as设置s is enabled, the browser transport now prefers watching Midjourney's existing in-page network traffic for the submitted job_id. It falls bac