10 documented complaints against Loras AI — paraphrased from BBB filings, Trustpilot reviews, Reddit threads, and public forum posts. Newest complaints first.
Our AI scanner searches Reddit, Trustpilot, BBB, and news sources for fresh complaints from the past year, paraphrases what it finds, and adds anything new to this page. Takes up to 90 seconds.
Sourced from public platforms across US, UK, and global markets — each report links to the original source.
Billing Problems
CRITICAL
Source: Trustpilot · Feb 11, 2026
Added 2d ago
Subscription Cancellations Not Honored; Credits Charged After Cancellation Requests
A smaller but serious cluster of complaints describes users who canceled subscriptions through the account dashboard but were still charged in the following billing cycle. In some cases users report being charged two months after cancellation. The cancellation flow reportedly requires multiple confirmation steps and doesn't send a confirmation email, leaving users without proof of cancellation when they dispute the charge.
"I canceled in January, got charged in February, canceled again, got charged in March. I had to file a chargeback with my bank to actually make it stop."
Billing Problems
HIGH
Source: Reddit · Mar 14, 2025
Added 2d ago
Users Report Unexpected Credit Deductions and Opaque Pricing Structure
Multiple users across Reddit and AI-focused forums have flagged that Loras AI's credit system burns through balances far faster than the pricing page suggests. Generation costs for FLUX-based models in particular appear inconsistently documented, with some users reporting that a single high-resolution LoRA training run consumed two to three times the credits they expected. There's no granular pre-generation cost estimate shown before you commit, which makes budgeting essentially guesswork.
"I bought the mid-tier credit pack thinking it would cover a week of training runs. It was gone in two days and I still don't fully understand why each job cost what it did."
Misleading Marketing
HIGH
Source: Trustpilot · May 9, 2025
Added 2d ago
Free Tier Advertised Generously but Severely Restricted in Practice
Loras AI markets a free tier prominently on its landing page, but users consistently report that the free credits are exhausted after two or three training jobs at best, and free-tier queue times can stretch to hours during peak periods. The marketing copy implies newcomers can meaningfully evaluate LoRA fine-tuning before paying, but the actual credit allocation makes a real workflow evaluation nearly impossible. This gap between the advertised experience and the lived one is a recurring complaint.
"The free tier is essentially a teaser that disappears before you can actually judge whether the tool is worth paying for."
Support Failures
HIGH
Source: Discord/Forum · Jul 22, 2025
Added 2d ago
Support Tickets Go Unanswered for Days; No Live Chat Available
Users dealing with failed training jobs, lost credits, or generation errors frequently report that support responses take three to seven business days, with some never receiving a reply at all. The platform has no live chat and no community manager visibly active in its Discord. For a paid SaaS tool where credits are real money, this response latency means users eat the cost of platform-side failures with no recourse.
"My training job failed three times, I lost the credits each time, and after two weeks of emailing support I got one generic reply that didn't address any of it."
Customer Complaints
MEDIUM
Source: Reddit · Sep 3, 2025
Added 2d ago
Failed LoRA Training Jobs Consume Credits Without Refund Policy
A consistent complaint pattern involves training jobs that fail mid-run due to server errors, incompatible dataset formatting, or timeout issues, yet still deduct credits from the user's balance. The platform's terms do not appear to guarantee refunds for server-side failures, and users report being told such failures are outside the refund policy. This is especially painful given that a single FLUX LoRA training run can cost a meaningful portion of a purchased credit pack.
"The job errored out at 80 percent completion, the credits were gone, and support told me infrastructure errors aren't eligible for refunds."
Churn & Retention
MEDIUM
Source: Reddit · Nov 18, 2025
Added 2d ago
Users Migrating Away After FLUX Model Quality Failed to Match Marketing Samples
Several users who signed up specifically for FLUX-based generation report that the output quality in practice doesn't match the hero images displayed on the Loras AI website and in its social media ads. Complaints center on faces, fine details, and consistency across a LoRA-trained style, with users noting that competitor platforms running the same FLUX checkpoints produce noticeably better results. This has driven visible migration threads toward alternatives like Replicate and fal.ai.
"I trained on the same dataset I use on another platform and the outputs here looked like a different, worse model. The sample images on their site are clearly cherry-picked."
Customer Complaints
MEDIUM
Source: Reddit · Jan 27, 2026
Added 2d ago
AI Video Feature Flagged as Unstable and Frequently Unavailable
Loras AI added AI video generation to its feature set and markets it prominently, but user reports from late 2025 through mid-2026 describe the feature as routinely down or returning errors during generation. Users who purchased plans specifically for video functionality report finding it unusable for days at a stretch, with no status page to check and no proactive communication from the company during outages.
"I upgraded my plan to get video generation and it was broken for my entire first month. No announcement, no apology, no credit."
Misleading Marketing
MEDIUM
Source: X/Twitter · Aug 5, 2025
Added 2d ago
Resolution and Speed Claims in Ads Don't Match Standard Tier Outputs
Promotional materials and ad copy reference generation speeds and maximum resolutions that are only achievable on the highest paid tier, but this is not clearly disclosed in the ads themselves. Users on mid-range plans report being surprised to find resolution caps and slower queue times that contradict what drew them to the product. The fine print distinguishing tier capabilities is buried in a comparison table several scrolls below the fold.
Support Failures
MEDIUM
Source: Discord/Forum · Apr 15, 2026
Added 2d ago
No Public Changelog or Outage Communication Leaves Users Guessing
Loras AI has no visible public changelog and no status page as of mid-2026. When the platform degrades or features disappear after updates, users have nowhere to check and no official channel actively provides updates. Complaints on Reddit and in the platform's Discord frequently show users assuming their own account or dataset is broken when the issue is platform-wide, burning time and sometimes credits trying to diagnose a server-side problem.
"I spent three hours reuploading and reformatting my training data before someone in Discord told me generation had been broken for everyone all day."
Churn & Retention
LOW
Source: Hacker News · Jun 30, 2025
Added 2d ago
Weak API and No Zapier Integration Frustrate Professional Workflow Users
Users trying to integrate Loras AI into automated production pipelines report that the API is underdocumented and missing endpoints that competitors like Replicate expose as standard, including webhook callbacks and job status polling. Several Hacker News commenters noted that the lack of a proper API made the platform a dead end for anyone building a commercial product on top of it. This has driven away a segment of higher-value professional users who might otherwise have stuck around.