TLDR: A Google AI Pro subscriber says a single Gemini video generation attempt consumed a full five hour usage cap in minutes, then blocked further use. Google acknowledged the complaint and is reviewing how its new compute based limits calculate credits.
Key Takeaways:
- Google AI plans moved to compute based Gemini limits that price prompts by complexity, features, and conversation length.
- A subscriber claims one failed video generation prompt burned the entire five hour allowance within minutes, despite the new credit style quota.
- If credit math overcharges edge cases, paying users may hit caps early, pushing more reliance on retries and limiting experimentation.
Paying for AI is supposed to buy time, not math. When compute credits behave oddly, the product feels like it runs out, not because you used it, but because it guessed wrong.
Paying for AI is supposed to buy time, not math. When compute credits behave oddly, the product feels like it runs out, not because you used it, but because it guessed wrong.
Q&A
What kind of prompt failures are most likely to trigger runaway compute credit charges?
Anything that forces repeated internal retries, model restarts, or heavy safety and validation steps can inflate compute use, especially during video generation.
How could compute based quotas change user behavior compared with fixed prompt caps?
Users may shorten prompts, avoid certain features, and split work across sessions to reduce the chance a single complex request spikes credits.
What would a fair fix look like if the cap is being charged too aggressively?
Google could cap credit burn per request, refund credits after identifiable failures, or smooth spikes by estimating cost before running the full pipeline.
Why might a video generation failure cost more than a successful attempt?
Failures can happen after the system has already invested compute in rendering, intermediate steps, or safety checks that still consume significant resources.
If Google confirms the issue, what broader impact could it have on AI plan pricing?
Persistent credit math complaints can pressure providers to standardize quotas, add usage transparency, or offer tiered limits that better match real user demand.
Top in AI
YouTube reshuffles AI labels to surface disclosure upfront
Nvidia Rubin CPX stalls as Groq integration tightens supply math
YouTube boosts AI labels as detection automation ramps
ElevenLabs brings Stan Lee voice, likeness back via licensing pact
No comments yet. Be the first to share your thoughts!