← Free, open source, non-profit← 免费、开源、非盈利

Bring your own key换成自己的 key

nanoMuse is free and non-profit; the model behind it costs money, and the developer pays for a starter pool — ¥10 per account, and ¥5 more for you and ¥5 for them each time a friend signs up with your code. When that is gone, the fastest way on is a key of your own. Below: where you are decides the provider — Alibaba Cloud Bailian in mainland China, OpenRouter everywhere else — then any other OpenAI-compatible provider. nanoMuse 免费、非盈利；背后的模型是要花钱的，开发者为每个账号垫付了一笔起步额度——¥10，每邀请一位新用户，你 +¥5、对方也 +¥5。用完之后，最快的办法是换成自己的 key。下面先看你在哪——中国大陆用阿里云百炼，海外用 OpenRouter——再是其他 OpenAI 兼容服务商的通用填法。

Where you are decides the provider先看你在哪

Bailian only signs up accounts with a mainland-China identity, and its international edition has neither model; OpenRouter is the easy way elsewhere — one account, one key, an international card, pay as you go. Both are presets in every client, the other one always a step away. Chat model and hands model are two settings: the chat model in the model picker, the hands model on its own row (phone: Settings → Hands; web and desktop: Connections); they default to deepseek-v4.1-flash and qwen3.8-27b on the allowance, and the presets pick the same two under your key. 百炼只给中国大陆身份注册，国际版也没有这两个模型；海外用 OpenRouter 最省事——一个账号一把 key，国际信用卡，按量付费。两家在每个客户端里都是预设，想换另一家也只差一步。对话模型和手的模型是两个设置：对话模型在模型选择里，手的模型单独一行（手机：设置 → Hands；网页版 / 桌面版：连接）；额度没用完时默认是 deepseek-v4.1-flash 和 qwen3.8-27b，换成自己的 key 之后预设也选这两个。

Alibaba Cloud Bailian (mainland China)阿里云百炼（中国大陆）

• A free quota for new accounts — the main models come with free tokens for a while after sign-up.新用户有免费额度——开通后主流模型都送一段时间的免费 token，先用起来不花钱。
• One key covers everything — chat (Qwen), pictures (qwen-image) and video (Wan) sit under the same key, so the avatar studio works too.一个 key 就够——对话（Qwen）、画图（qwen-image）、视频（Wan）都在同一个 key 下，形象生成也能直接用。
• Direct from China, OpenAI-compatible — nanoMuse already knows the endpoint; you only paste the key.国内直连，OpenAI 兼容——nanoMuse 里已经预置好了地址，只需要贴 key。

Five steps五步开通

1. Open Bailian.开通百炼。 Go to bailian.console.aliyun.com and sign in with an Alibaba Cloud account (create one if needed; identity verification is required). On first entry, accept Enable Model Studio — it is free. 打开 bailian.console.aliyun.com，用阿里云账号登录（没有的话注册一个，需要实名认证）。第一次进入会提示「开通百炼服务」，点开通即可，不收费。
2. Create an API key.创建 API key。 Open the API-KEY page (avatar, top right → API-KEY), click Create my API-KEY, keep the default workspace, confirm. 进入 API-KEY 管理页（右上角头像 → API-KEY），点「创建我的 API-KEY」，归属选默认业务空间，确定。
3. Copy it.复制 key。 The key starts with sk-. Click View, then Copy. It is for nanoMuse only — never send it to anyone or paste it into a chat. 新建的 key 以 sk- 开头。点「查看」再「复制」。这串字符只给 nanoMuse 用，不要发给任何人、不要贴到聊天里。
4. Paste it into nanoMuse.贴到 nanoMuse 里。
   • Phone: Set it up on the allowance card opens the provider form pre-filled for Bailian; or Settings → Providers → Add → OpenAI, base URL https://dashscope.aliyuncs.com/compatible-mode with /v1 on. Paste the key, save.手机：额度卡片上的「去设置」会直接打开预填好的「阿里云百炼」表单；或者 设置 → 服务商 → 添加 → OpenAI，地址填 https://dashscope.aliyuncs.com/compatible-mode（勾上 /v1）。贴 key，保存。
   • Web / desktop: Set it up on the allowance card opens Connections with Alibaba Cloud Bailian chosen; or Connections → Model → Alibaba Cloud Bailian. Paste the key, save.网页版 / 桌面版：额度卡片上的「去设置」会打开「连接」页并选好「阿里云百炼」；或者 连接 → 模型 → 阿里云百炼。贴 key，保存。
5. Pick a model, say something.选一个模型，试一句。 After saving, nanoMuse lists the models the key can use; take qwen3.8-flash (cheap and quick) or qwen3.8-27b (stronger) for chat. Back in the chat, one answered sentence means it works. For the avatar, choose qwen-image-3.0 and wan2.2-i2v-flash under Connections → Image & video models. 保存后 nanoMuse 会自动列出这个 key 能用的模型；对话选 qwen3.8-flash（便宜快）或 qwen3.8-27b（更强）。回到聊天说一句话，能回答就成了。想让它换形象，再到 连接 → 图像与视频模型 里选 qwen-image-3.0 和 wan2.2-i2v-flash。

What it costs afterwards之后怎么算钱

Past the free quota, Bailian bills at its list prices (September 2026, Beijing region): qwen3.8-27b ¥3 in / ¥12 out per million tokens, qwen3.8-flash ¥0.8 / ¥2.7, qwen-image-3.0 ¥0.18 a picture, wan2.2-i2v-flash ¥0.10 a second. An ordinary day of chatting is a few fen; set a usage alert in the Bailian console. nanoMuse itself charges nothing. 免费额度用完后按百炼的标价计费（2026 年 9 月，北京地域）：qwen3.8-27b 每百万 token 输入 ¥3 / 输出 ¥12，qwen3.8-flash ¥0.8 / ¥2.7，qwen-image-3.0 每张 ¥0.18，wan2.2-i2v-flash 每秒 ¥0.10。正常聊一天几分钱；建议在百炼控制台设一个用量告警。nanoMuse 自己不收任何费用。

OpenRouter (everywhere else)OpenRouter（海外）

• One account, one key, hundreds of models behind it; an international card, pay as you go.一个账号、一把 key，背后有几百个模型；国际信用卡付费，按量扣。
• No proxy, no mainland identity.不用代理，不用国内身份。
• The phone apps sign in to OpenRouter with one tap (Android, iPhone) — no key to copy.手机 App 可以一键登录 OpenRouter（Android、iPhone），不用手抄 key。

1. Sign up. Open openrouter.ai and sign in with Google, GitHub or an e-mail.注册。打开 openrouter.ai，用 Google、GitHub 或邮箱登录。
2. Add credits. Avatar → Credits → Add Credits; five dollars lasts a long while.充值。右上角头像 → Credits → Add Credits，国际信用卡；先充 5 美元足够用很久。
3. Make a key. Avatar → Keys → Create Key, name it (nanoMuse, say), copy the string starting sk-or-v1-. For nanoMuse only; never send it to anyone.拿 key。头像 → Keys → Create Key，起个名（比如 nanoMuse），复制以 sk-or-v1- 开头的那串。只给 nanoMuse 用，别发给任何人。
4. Paste it into nanoMuse. Phone: Settings → Providers → Add → OpenRouter → Sign in, approve in the browser and the key comes back by itself — or paste it into the OpenRouter form. Web and desktop: Connections → Models → OpenRouter, paste, save; the address is filled in.贴到 nanoMuse 里。手机：设置 → 服务商 → 添加 → OpenRouter → 「登录」，浏览器里批准一下，key 自动回到 App，不用复制；或者把 key 贴进 OpenRouter 表单。网页版 / 桌面版：连接 → 模型 → 「OpenRouter」，贴 key，保存。地址已经填好。

Pick deepseek/deepseek-v4.1-flash for the chat and qwen/qwen3.8-27b for the hands. Each model's price is on its OpenRouter page; Settings → Limits puts a monthly cap on the key. 对话选 deepseek/deepseek-v4.1-flash，手的模型选 qwen/qwen3.8-27b。每个模型的单价在 OpenRouter 的模型页上写着；Settings → Limits 可以给 key 设每月上限。

Any other OpenAI-compatible provider其他 OpenAI 兼容服务商

Anything that speaks the OpenAI API works: DeepSeek, Zhipu, Moonshot, OpenRouter, OpenAI itself, or your own vLLM / Ollama. The generic fill-in: 任何 OpenAI 兼容接口都能填：DeepSeek、智谱、月之暗面、OpenRouter、OpenAI 本身，或你自己跑的 vLLM / Ollama。通用填法：

The web and desktop Connections page has presets for the common providers: choose one and the address is filled in — only the key is missing. 网页版和桌面版的「连接」页里有常见服务商的预设，选中即填好地址，只差 key。