Issues

• [bug]: first load after an update shows raw translation keys (service worker serves stale locales)

  invoke-ai/InvokeAI#9908 · Filter by author Pfannkuchensack opened on Oct 18, 2026
  1

• [bug]: model cache leaves the requested model partly resident after splitting the last offload candidate

  invoke-ai/InvokeAI#9907 · Filter by author Pfannkuchensack opened on Oct 18, 2026

• [bug]: Reject unresolvable MiniMax H3 LoRA targets

  bug
  invoke-ai/InvokeAI#9904 · Filter by author JPPhoto opened on Oct 17, 2026
  

• [bug]: Diffusers repo-id installs never apply the model_index.json component filter

  bugmodel manager
  invoke-ai/InvokeAI#9901 · Filter by author Pfannkuchensack opened on Oct 17, 2026

• MiniMax H3: support PEFT DoRA LoRAs (lora_magnitude_vector) on full bf16 transformers

  invoke-ai/InvokeAI#9900 · Filter by author lstein opened on Oct 17, 2026

• [enhancement]: Flux.3

  enhancement
  invoke-ai/InvokeAI#9885 · Filter by author CLG68 opened on Oct 16, 2026
  1

• [enhancement]: Drag & Dropo Files

  enhancement
  invoke-ai/InvokeAI#9882 · Filter by author Quasi-Social opened on Oct 16, 2026
  1

• [bug]: SDXL images distorted after macOS 27 update

  bug
  invoke-ai/InvokeAI#9880 · Filter by author painellope opened on Oct 16, 2026
  2

• Recall shortcuts not doing anything

  invoke-ai/InvokeAI#9874 · Filter by author joshistoast opened on Oct 15, 2026

• Manual 'render at' needs to lock to aspect ratio

  invoke-ai/InvokeAI#9873 · Filter by author joshistoast opened on Oct 15, 2026

• Controlnet undefined to value validation

  invoke-ai/InvokeAI#9872 · Filter by author joshistoast opened on Oct 15, 2026

• [bug]: MiniMax H3 single-file text encoder silently drops the cpu_only setting

  bug
  invoke-ai/InvokeAI#9866 · Filter by author Pfannkuchensack opened on Oct 15, 2026
  

• [bug]: Switch for running MiniMax H3 Text Encoder on CPU resets itself when used

  bug
  invoke-ai/InvokeAI#9865 · Filter by author AuralVertigo opened on Oct 15, 2026
  

• [bug]: webv2 resubmits a run indefinitely after a 500 on enqueue; the run is hidden from the Queue panel and survives reloads

  7.0.0
  Bug

  invoke-ai/InvokeAI#9854 · Filter by author unmarco opened on Oct 14, 2026
  1

  

• [Enhancement]: Different image formats and better compression

  enhancement
  invoke-ai/InvokeAI#9850 · Filter by author Deniryer opened on Oct 14, 2026
  

• [bug]: sibling worker can consume a GPU borrow-release wake-up

  invoke-ai/InvokeAI#9849 · Filter by author wunianze666-netizen opened on Oct 14, 2026
  

• [bug]: Qwen3-VL-32B-Int8 Text Encoder isn't recognized as MiniMax-H3 TE component

  bug
  invoke-ai/InvokeAI#9846 · Filter by author NiceDragon8066 opened on Oct 14, 2026
  1

  

• [bug]: Interrupted model download restarts from the beginning for download retry option

  bug
  invoke-ai/InvokeAI#9841 · Filter by author NiceDragon8066 opened on Oct 13, 2026
  2

  

• [bug]: Wan-2.2 A14B High and Low Noise models cannot be loaded simultaneously. Standard workflows are missing.

  bug
  invoke-ai/InvokeAI#9840 · Filter by author NiceDragon8066 opened on Oct 13, 2026
  

• [enhancement]: Need support for NVFP4 text encoder Qwen3-VL-32B for MiniMax-H3

  enhancement
  invoke-ai/InvokeAI#9839 · Filter by author NiceDragon8066 opened on Oct 13, 2026
  

• [bug]: Wan 2.2 A14B at 720p does not fit 24 GB (no working-memory estimate); SageAttention makes the overflow slower

  bug
  invoke-ai/InvokeAI#9831 · Filter by author Pfannkuchensack opened on Oct 12, 2026
  

• [enhancement]: Bump PyPatchMatch to latest

  enhancement
  invoke-ai/InvokeAI#9826 · Filter by author iwr-redmond opened on Oct 12, 2026
  

• [enhancement]: Adjustment layers: Vibrance/Saturation and Clarity/Dehaze

  enhancement
  invoke-ai/InvokeAI#9813 · Filter by author shanedk opened on Oct 11, 2026
  5

  

• [enhancement]: Add "Keep Input Language" Option to Prompt Expander (with Auto-Translation on Generation)

  enhancement
  invoke-ai/InvokeAI#9812 · Filter by author baramofme opened on Oct 11, 2026
  3

  

• [enhancement]: Load torchao-quantized single-file transformers (unsloth FP8)

  enhancement
  invoke-ai/InvokeAI#9798 · Filter by author Pfannkuchensack opened on Oct 10, 2026