A Mage.space alternative with subscription optional, an open API, and credits that do not expire

Everything this page cites from Mage, its FAQ included, comes from one page: its membership page, read on 11 September 2026 (mage.space/pricing now redirects there). Mage is subscription-first: monthly membership tiers, with membership Gems that its FAQ says expire after 30 days, alongside Gems bought for pay-per-generation use. Its membership table caps video by tier, with no video on the entry tier. Mage names Stripe as its billing partner on that page; we opened no account, so that describes that page on that date rather than its logged-in purchase flow.

Against that, Vellria's pay-as-you-go credits carry no membership and nothing on a timer by default, though an optional monthly plan is available if you would rather have one. Credits land once a payment on our partner's checkout goes through, by card or by another option your country offers, and the balance moves only when you start a run. Gems that lapse after 30 days have no counterpart here: a balance left alone in March is still there in May. Nor is video held behind a tier. Every image and video model opens to any account holding the credits, and each one's price sits on its own page and on the pricing page.

Refunds differ too. Mage states that it typically does not offer refunds, citing the unlimited nature of its plans. Our rules are narrower and written down in the refund policy: a run that fails gives its credits back without anyone asking, and for credits you never used, you can ask for your money back from the console's support page, where each request is judged on its own.

Mage's FAQ states that automation is strictly forbidden on Mage, and tells business users to contact it about API access. The API itself is not public: Mage's own FAQ calls it a premium offering in limited beta and says "We are currently at capacity", and no api., docs. or developer. subdomain resolved when we checked on 11 September 2026. Each model page, for example Seedream 5 Lite or Qwen 3 Pro, shows the id and the sizes to send, and the NSFW image API page has the request shape.

On Vellria, scripting the service is the intended use rather than a breach. You make a key in the console yourself, with no form and no queue, put it in the Authorization header, and send a model id, a prompt and that model's options. The run starts asynchronously; your script keeps the generation id and checks back until the result is ready. The guard rails are published per-minute limits instead of a ban, and a request over the limit gets a 429 whose Retry-After header tells your code how long to wait. The API docs have the shapes, and the Civitai alternative sets the same key-and-balance setup beside a platform whose API is open to scripts.