I subscribed to Galaxy AI (now Magica AI) in mid‑December 2025 on the $14.99 monthly plan and canceled in the second week of June 2026. During this period, I accumulated close to 70 million generation credits and used the service extensively.
From the first weeks, many outputs were seriously flawed and did not match my prompts in any reasonable way. I repeatedly saw broken results (including obvious visual defects and unusable generations) while the system still consumed credits as if the outputs were successful. Over time, this added up to tens of millions of credits lost on failed or unusable generations.
I followed the platform’s own instructions and contacted support multiple times by email to ask for restoration of the credits that were burned on clearly defective results. I never received a response to any of these messages, even though I can document that they were sent. Eventually, support acknowledged only a fraction of the lost value by offering 15 million credits back against well over 60 million credits consumed on failed generations. In my view, this is not a good‑faith resolution of a documented service failure.
While this was happening, Galaxy AI rebranded to Magica AI. In my experience, the rebrand coincided with a degraded interface, the loss or reduction of useful tools, and no meaningful improvement in reliability. The system continued to make the same mistakes, apologize, and then repeat the same issues again, consuming more credits each time.
Because I wanted to understand whether my case was isolated, I checked public feedback on third‑party sites, including G2 and TrustPilot and other review platforms, as well as posts on forums and social media. I found multiple complaints from other users describing similar patterns: credit consumption on broken output, poor or absent support responses, and frustration with the rebrand and product changes. This reinforced my impression that my experience was not a one‑off incident but part of a broader pattern of issues with the service.
I have kept records of my subscription dates, credit usage, failed generations, and support attempts, and I am prepared to share this documentation with relevant authorities if needed. This review reflects my genuine experience as a paying customer over roughly six months of continuous use. Based on this experience, I do not recommend Magica AI. In my case, the Magica + Claude setup failed to produce reliable, usable output for my projects, despite consuming large amounts of credits and repeatedly apologizing for the same issues. Users who rely on consistent, production‑quality output should be especially cautious. Review collected by and hosted on G2.com.
The support with Galaxy AI leaves something to be desired in that if the AI Support cannot help you and it tells you to email Support, you might now ever receive a reply email. I know I haven't! Review collected by and hosted on G2.com.
