Common Issues

Safe checks for startup, licensing, downloads, rendering and support

Last updated: 2026-10-27

Start with the message shown in the app and the exact action that failed. Keep your original images and important work backed up. Avoid deleting model, project or render folders as a general troubleshooting step.

App Won't Start

Record your app version, Windows version and what happens at launch. Check System Requirements and use the supported Windows installer. If Windows reports a missing runtime or driver, retain the exact message for support rather than making unrelated system changes or running the app as administrator by default.

If the interface opens but the backend or GPU is unavailable, save the current-session diagnostics from Send Feedback…. A saved report does not require a successful online submission. See Report an issue.

License Activation Errors

  • Activation limit reached: manage your registered computers from your account and release an unused slot. A removed device must activate again; this does not erase its local work.
  • Invalid key: copy the full key from your account or purchase message. Never post the key in a public report.
  • Server unreachable / pending verification: check connectivity and try online verification again. A never-verified key does not unlock first-time offline rendering.
  • Revalidation required: reconnect; rendering's offline window is seven days after a successful check.
  • Revoked, suspended or device removed: follow the actual message and activation guide. Do not delete device identity files to bypass it.

Model Download Failures

Use the model catalogue for the selected pipeline. Check its actual download, installation and readiness state before starting another operation.

  1. Confirm enough free space on the selected storage drive for that model and its installation, not a fixed universal allowance.
  2. Read the error. Gated models may require accepting the model host's terms and supplying your own access token through the app's authentication settings.
  3. Retry using the model's download/repair action when offered. An interrupted download is not proof the installation is ready; wait for the app's final state.
  4. If readiness remains uncertain, retain the model name, selected precision and error for support. Do not manually delete shared weights or caches while work is using them.

Processing Issues

Out of Memory Error

Stop submitting more work. Wait for the current job's settled status and check GPU Management. Choose a model/precision and processing size that fit your hardware. CPU offload, where supported, trades GPU pressure for system RAM and slower work; it is not a universal fit guarantee.

More template faces alone do not establish lower memory use. Keep a small first task and use the selected pipeline's defaults rather than copying settings from a different model family.

Processing Stalls

Inspect the existing job and its progress/error state before starting again. A delayed response can leave an operation still running. Use its supported Cancel control and wait for a terminal result; cancellation requested is not proof that GPU work has stopped.

If the app was interrupted, reopen the existing render/sequence and inspect its recovery options and completed output. Resume is available only for supported jobs; do not assume every tool or generation can resume. Keep error details if the state remains unknown, and avoid repeatedly launching duplicate work.

Follow Rendering and Recovery for queue states, Stop confirmation and the different recovery paths for exports, sequences, model downloads and uploads. For an unexpectedly empty library, check the configured drive, active Project and Storage fallback warning before cleanup; see Storage and Backup.

Seams Visible in Output

Check that the input projection, selected template, tile placement and prompts match the intended panorama. Compare a lower-strength test and inspect both poles and tile boundaries. Settings that help one pipeline may harm another. Factory templates are chosen in Templates; there is no Settings → Templates download remedy.

Poor Quality Results

Compare the input, model defaults and a single controlled change. Denoising strength determines how far the image changes; steps and guidance behave differently across models. More steps are not always better, especially with distilled/turbo schedules. See Pipelines and Directional Prompts.

GPU Not Detected

Check the GPU shown by the app against System Requirements. A supported NVIDIA CUDA device and compatible driver are needed for GPU processing. Record the displayed device/driver/error details; do not infer compatibility from VRAM alone.

Error Messages

  • Template not found: select an available factory or saved template in the template picker; retain the missing template's name if a saved Workflow references it.
  • Invalid image dimensions: check the import tool's accepted projection and dimensions. An ordinary 360 panorama is 2:1 equirectangular; use Format Converter for other supported projections.
  • Model not found: choose an installed compatible model or complete its catalogue download. Changing pipelines can also change the required add-ons.
  • Project busy: finish or cancel active work and wait for it to settle before changing roots. See Projects.

Confirm the separate Settings → Gallery connection and account permission, not just desktop activation. Check the chosen source, dimensions and returned publishing status. Upload completion, server processing and a public-ready item are different stages. Inspect an existing item before retrying after a lost response. See Gallery Publishing.

Report an issue

  1. Open the app's Help menu and choose Send Feedback….
  2. Choose the report type and describe the action, expected result, actual result and any displayed error. Do not include keys, tokens or private material in the message.
  3. Diagnostics are optional and off initially. Save diagnostics to file retains current-session evidence. Review it before sharing; privacy filtering cannot guarantee arbitrary error text contains no sensitive information.
  4. Send Feedback first saves a local report under Documents/360 Hextile/Feedback, then attempts delivery. If saving fails, the draft remains and Copy message is available; no report was sent.
  5. If diagnostic preparation fails, use Send without diagnostics or try preparing again. You can keep the locally saved report for a later support message.
  6. A confirmed receipt ID establishes delivery to the feedback intake. Save that ID. A timeout or uncertain response can mean the server received the report even though the app cannot confirm it. Inspect the result before an explicit retry, which may create a duplicate.

Each explicit send/retry saves another local report. There is no automatic later upload from that folder. A local save, a receipt and a support email reply are separate outcomes.

Getting More Help

Use docs search, Contact, or the community. Share the app version, affected tool/model, exact error and a small reproduction, plus reviewed diagnostics if useful. Keep public posts free of private files and credentials.

Esc
Searching...
No results found.
Type to search the documentation
↑↓ Navigate ↵ Select Esc Close

We use cookies for analytics to understand how visitors use our site. Privacy Policy