When something goes wrong during a generation, it can be hard to know whether it's a Layer issue or a problem with the underlying model provider, like Kling, Gemini, or Veo. This article helps you identify what's happening and take the right action quickly.
Is it Layer, or the model provider?
Layer connects you to a range of third-party AI models, including Kling (Tencent), Gemini and Veo (Google), Flux, Qwen, and others. When errors occur, the cause can sit with either Layer's infrastructure or the model provider's API. The two behave differently, so it's worth knowing how to tell them apart.
Signs it's a model provider issue
The error appears only on one specific model (e.g. Gemini 3 Pro, Kling 3, Veo 3.1) while others continue working normally
You see messages like "A service we rely on is unavailable. Try again shortly."
Generations on that model are extremely slow, timing out, or silently failing
Switching to a different model for the same task works fine
Other users on the same model are reporting the same issue
Signs it's a Layer issue
Errors appear across multiple models or modalities at the same time
The platform itself is slow, unresponsive, or showing a black screen
Core features like uploading assets, opening sessions, or navigating the workspace are broken, not just generation
Our Status Page shows an active incident
What to do
If it's a model provider issue
The most effective thing you can do is switch to an alternative model while the provider resolves their issue. Layer gives you access to many models across each modality, so you're rarely blocked entirely.
Recommended alternatives by use case:
If this model has issues | Try this instead |
Gemini / Nano Banana (image generation and prompt editing) | Filter by latest models via Models tab on the app for generations. You can use 'New' label to understand which models are the latest. |
Kling 3 or Veo (video) | Filter by latest models via Models tab on the app for generations. |
|
|
Provider issues are outside of Layer's control, but our team monitors them closely and communicates updates through our Status Page. Enterprise customers will also receive updates through their dedicated Slack support channel.
If it's a Layer issue
Check our Status Page first, as active incidents are listed there in real time and our team posts updates as they become available.
If no incident is listed but you're still experiencing issues, try the following before contacting support:
1 | Hard refresh the page. Hold Shift and click Refresh, or use Ctrl+Shift+R (Windows) / Cmd+Shift+R (Mac). |
2 | Clear your browser cache for layer.ai. |
3 | Try a different browser or open an incognito / private window. |
4 | Navigate directly to your workspace URL (e.g. app.layer.ai/yourworkspace) and do a hard reload. |
If the issue persists after these steps, contact support with your session link so our team can investigate directly.
Do failed generations cost Creative Units?
No. If a generation fails or gets stuck without completing, Creative Units are not deducted from your balance. You are only charged for generations that finish successfully.
If you believe credits were consumed during a failed generation, you can verify this by going to Workspace > Usage and reviewing your usage events. Each logged event corresponds to a completed generation. If something still looks off after checking, contact support with the relevant session link.
Frequently Asked Questions
How do I know which model caused the error?
Look at which model was selected when the error occurred. If only that model is affected and others work normally, it's almost certainly a provider-side issue. Try switching models to confirm.
Will Layer notify me about outages?
Active incidents and resolutions are posted to our Status Page at status.layer.ai. Enterprise customers will also receive updates through their dedicated Slack channel when their workflow is impacted.
The error message says "Something went wrong." How do I find out more?
Check the Status Page for any active incidents. If nothing is listed, try switching to a different model first. If the problem persists across models, share your session link with support so we can pull the logs and investigate.
Gemini / Kling / Veo has been down for a long time. Can Layer speed up the fix?
Provider outages are resolved by the model provider, not Layer. Our team stays in direct contact with providers during extended outages and will communicate when things stabilize. In the meantime, we recommend building familiarity with alternative models so your workflow isn't dependent on a single provider.
I switched models but the quality isn't what I need. What are my options?
Different models have different strengths. If you have a trained LoRA, try applying it with an alternative base model. Our team is also happy to suggest the best available alternative for your specific use case; just reach out via support.
My generations were slow but eventually completed. Were CUs charged?
Yes. If a generation completes successfully, CUs are charged as normal regardless of how long it took. CUs are only not charged when a generation fails to complete entirely.
Still experiencing issues? Contact our support team with your session link and a description of the error and we'll investigate.