Short answer
MiniMax released H3 Image (h3-image) on November 11, 2026: an open-weights model for text-to-image and image editing. Native 4096x4096 (about 16 megapixels), MiniMax's own figure of under 10 seconds per image, up to 16 reference images in one pass, and the weights are published.
Whether you can use it is not the question — which route you take is. MiniMax offers two entry points:
- Cloud, via Hailuo AI: the launch material states plainly that this is a paid feature. It does not state a per-image price, a free allowance, the accepted payment methods, or which regions it serves.
- Local, by deploying the weights: this needs no account and no card at all, only a GPU that can handle it. ComfyUI 0.45.0 or newer is required, and the template package needs 0.11.104 or newer.
So the answer to "what do I do without an overseas account or an overseas card" is: run it locally. That is also the most practical part of the open-weights release for users in China and Russia — it removes the "first get an account that can pay" precondition entirely.
h3-image is not on TeamoRouter yet. Once it is listed you will not need a Hailuo account or an overseas card: use the TeamoRouter key you already have and set model to h3-image. Until then, on the text side you can use minimax-m3.1-flash or MiniMax-M3 — those are language models and will not draw anything for you.
Where the official route actually stops you
Separate what the launch material says from what it leaves out.
Stated: the weights are open and can be self-hosted; ComfyUI has a ready-made template; cloud use runs through Hailuo AI and is a paid feature.
Not stated: the per-image price in the cloud, whether a free allowance exists, concurrency and rate limits, which payment methods Hailuo AI accepts, and which countries or regions it serves.
Those last items are exactly what someone without an overseas account or card needs to know. We are not filling in those numbers for MiniMax. What is certain: the cloud route requires solving payment first, and the local route does not.
Three routes and what each demands
| Route | GPU | Cost | Overseas account / card | Usable today |
|---|---|---|---|---|
| Hailuo AI (official cloud) | No | Yes — MiniMax calls it a paid feature | Payment methods not published | Yes, but paid |
| Local weights | Yes | Not billed per image | No | Yes |
| TeamoRouter (once listed) | No | Billed against your balance by usage | No | Not yet |
The local route costs you a GPU and time. MiniMax's figure is 10 seconds per image; real local speed depends on your card. If the official weights will not fit your VRAM and RAM, find a quantised build suited to your GPU rather than forcing it. Picking the right build is a good job to hand to an agent that can read your hardware config — faster than trial and error.
What you can move forward without an account
The generation step itself has no stand-in, but the work around it can start now:
- Get the prompts right. The most useful lesson from the official demos: name concrete objects — metal, glass, ground dust, stitching at a collar, skin texture, cracked makeup — rather than writing "cinematic" and hoping.
- Settle the numbering scheme. Up to 16 reference images per pass, with people, outfits and colour cards supplied separately and tied together by number ("person 3, palette 5"). Sorting that mapping out in advance saves a lot of rerolls.
- Build the batch list and scripts. When the weights are running or
h3-imageis listed, you plug them straight in.
All three can go to minimax-m3.1-flash or MiniMax-M3. Both are already on TeamoRouter and work with the same key; reach for MiniMax-M3 when the text needs to be more careful or the prompt engineering more involved.
How to call it once it lands
When h3-image enters our catalog, neither your key nor the base URL changes — only model:
from openai import OpenAI
client = OpenAI(base_url="https://api.teamorouter.com/v1", api_key="YOUR_KEY")
resp = client.images.generate(
model="h3-image",
prompt="Place these nine pieces of furniture in the empty room, state where each goes, keep the existing decor",
size="4096x4096",
)
print(resp.data[0].url)
h3-image is not on TeamoRouter yet. Once it is live, the same key and model: h3-image is all you change; until then, use minimax-m3.1-flash.
How to pay
No subscription: top up your balance in the console from $5 and each request is billed against it by actual usage. Cards are supported, and so are USDT, USDC and ETH. The same key also calls Claude and GPT models.
New accounts start with a $0 balance — top up before your first request. Top up from $5. If a request returns 402 (400 on the OpenAI format) with "wallet balance is insufficient", your balance is empty — your base URL and key are fine. Top up and retry.
FAQ
Can I use H3 Image directly from China?
MiniMax gives two routes: Hailuo AI in the cloud, which is a paid feature, or download the weights and run them yourself. It has not published which payment methods or regions Hailuo AI supports, so there is no official answer to "can I just pay from here". If you would rather skip accounts and payment entirely, the weights are open and the local route needs neither.
Does it work from Russia?
Same logic: the cloud route sits behind Hailuo AI and is paid, with no published list of accepted payment methods. Local deployment depends on no account and no card, and the weights are already available — that is the ready fallback.
What if I have no overseas credit card?
Run the weights locally: no account, no card, the cost is a GPU and time. Or wait for h3-image on TeamoRouter and pay by usage with our key. Today you can already use minimax-m3.1-flash to prepare prompts and batch jobs.
Can I use it on TeamoRouter now?
Not yet — h3-image has not entered our catalog. Once it does, no Hailuo account and no overseas card are needed: same key, new model ID.
What do I need for a local setup?
ComfyUI 0.45.0 or newer, and template package 0.11.104 or newer. If VRAM or RAM is short, use the quantised build that fits your card instead of the official weights.
Once it is live, is it just a model-name swap?
Yes. Same key, same base URL, only model changes.
Related reading
- MiniMax H3 Image is out: model ID, price and how to call it: the specs, the current state of unpublished pricing, and both calling formats
- MiniMax H3 Image vs other image models: the head-to-head comparison