[bug]: NSFW checker not working, produces an error "Unrecognized model"

[magnusviri]

Is there an existing issue for this problem?

• I have searched the existing issues

Install method

Manual

Operating system

macOS

GPU vendor

Apple Silicon (MPS)

GPU model

M4 Pro

GPU VRAM

48GB

Version number

v6.8.0rc1

Browser

Brave 1.81.137 (Official Build) (arm64) Chromium: 139.0.7258.158

System Information

{
"version": "6.8.0rc1",
"dependencies": {
"absl-py" : "2.3.1",
"accelerate" : "1.10.1",
"annotated-types" : "0.7.0",
"anyio" : "4.10.0",
"attrs" : "25.3.0",
"babel" : "2.17.0",
"backrefs" : "5.9",
"bidict" : "0.23.1",
"blake3" : "1.0.6",
"blessed" : "1.22.0",
"cattrs" : "25.2.0",
"certifi" : "2025.8.3",
"cffi" : "2.0.0",
"cfgv" : "3.4.0",
"charset-normalizer" : "3.4.3",
"click" : "8.2.1",
"codefind" : "0.1.7",
"colorama" : "0.4.6",
"coloredlogs" : "15.0.1",
"compel" : "2.1.1",
"contourpy" : "1.3.3",
"coverage" : "7.10.6",
"CUDA" : "N/A",
"cycler" : "0.12.1",
"Deprecated" : "1.2.18",
"diffusers" : "0.33.0",
"distlib" : "0.4.0",
"dnspython" : "2.8.0",
"dynamicprompts" : "0.31.0",
"einops" : "0.8.1",
"Faker" : "37.8.0",
"fastapi" : "0.116.2",
"fastapi-events" : "0.12.2",
"filelock" : "3.19.1",
"flatbuffers" : "25.2.10",
"fonttools" : "4.60.0",
"fsspec" : "2025.9.0",
"gguf" : "0.17.1",
"ghp-import" : "2.1.0",
"gitdb" : "4.0.12",
"GitPython" : "3.1.45",
"gprof2dot" : "2025.4.14",
"griffe" : "1.14.0",
"h11" : "0.16.0",
"hf-xet" : "1.1.10",
"httpcore" : "1.0.9",
"httptools" : "0.6.4",
"httpx" : "0.28.1",
"huggingface-hub" : "0.35.0",
"humanfriendly" : "10.0",
"humanize" : "4.12.1",
"identify" : "2.6.14",
"idna" : "3.10",
"importlib_metadata" : "8.7.0",
"iniconfig" : "2.1.0",
"invisible-watermark" : "0.2.0",
"InvokeAI" : "6.8.0rc1",
"jax" : "0.7.1",
"jaxlib" : "0.7.1",
"jedi" : "0.19.2",
"Jinja2" : "3.1.6",
"jurigged" : "0.6.1",
"kiwisolver" : "1.4.9",
"lsprotocol" : "2023.0.1",
"Markdown" : "3.9",
"MarkupSafe" : "3.0.2",
"matplotlib" : "3.10.6",
"mediapipe" : "0.10.14",
"mergedeep" : "1.3.4",
"mkdocs" : "1.6.1",
"mkdocs-autorefs" : "1.4.3",
"mkdocs-get-deps" : "0.2.0",
"mkdocs-git-revision-date-localized-plugin": "1.4.7",
"mkdocs-material" : "9.6.20",
"mkdocs-material-extensions" : "1.3.1",
"mkdocs-redirects" : "1.2.2",
"mkdocstrings" : "0.30.0",
"mkdocstrings-python" : "1.18.2",
"ml_dtypes" : "0.5.3",
"mpmath" : "1.3.0",
"mypy" : "1.18.2",
"mypy_extensions" : "1.1.0",
"networkx" : "3.5",
"nodeenv" : "1.9.1",
"numpy" : "1.26.4",
"onnx" : "1.16.1",
"onnxruntime" : "1.19.2",
"opencv-contrib-python" : "4.11.0.86",
"opt_einsum" : "3.4.0",
"ovld" : "0.5.13",
"packaging" : "25.0",
"paginate" : "0.5.7",
"parso" : "0.8.5",
"pathspec" : "0.12.1",
"picklescan" : "0.0.31",
"pillow" : "11.3.0",
"platformdirs" : "4.4.0",
"pluggy" : "1.6.0",
"polyfactory" : "2.19.0",
"pre_commit" : "4.3.0",
"prompt_toolkit" : "3.0.52",
"protobuf" : "4.25.8",
"psutil" : "7.1.0",
"pudb" : "2025.1",
"pycparser" : "2.23",
"pydantic" : "2.11.9",
"pydantic-settings" : "2.10.1",
"pydantic_core" : "2.33.2",
"pygls" : "1.3.1",
"Pygments" : "2.19.2",
"pymdown-extensions" : "10.16.1",
"pyparsing" : "3.2.4",
"PyPatchMatch" : "1.0.2",
"pytest" : "8.4.2",
"pytest-cov" : "7.0.0",
"pytest-datadir" : "1.8.0",
"pytest-timeout" : "2.4.0",
"python-dateutil" : "2.9.0.post0",
"python-dotenv" : "1.1.1",
"python-engineio" : "4.12.2",
"python-multipart" : "0.0.20",
"python-socketio" : "5.13.0",
"pytz" : "2025.2",
"PyWavelets" : "1.9.0",
"PyYAML" : "6.0.2",
"pyyaml_env_tag" : "1.1",
"regex" : "2025.9.18",
"requests" : "2.32.5",
"requests-testadapter" : "0.3.0",
"ruff" : "0.11.13",
"ruff-lsp" : "0.0.62",
"safetensors" : "0.6.2",
"scipy" : "1.16.2",
"semver" : "3.0.4",
"sentencepiece" : "0.2.0",
"setuptools" : "80.9.0",
"simple-websocket" : "1.1.0",
"six" : "1.17.0",
"smmap" : "5.0.2",
"snakeviz" : "2.2.2",
"sniffio" : "1.3.1",
"sounddevice" : "0.5.2",
"spandrel" : "0.4.1",
"starlette" : "0.48.0",
"sympy" : "1.14.0",
"tokenizers" : "0.22.0",
"torch" : "2.7.1",
"torchsde" : "0.2.6",
"torchvision" : "0.22.1",
"tornado" : "6.5.2",
"tqdm" : "4.67.1",
"trampoline" : "0.1.2",
"transformers" : "4.56.1",
"typing-inspection" : "0.4.1",
"typing_extensions" : "4.15.0",
"tzdata" : "2025.2",
"urllib3" : "2.5.0",
"urwid" : "3.0.3",
"urwid_readline" : "0.15.1",
"uvicorn" : "0.35.0",
"uvloop" : "0.21.0",
"virtualenv" : "20.34.0",
"watchdog" : "6.0.0",
"watchfiles" : "1.1.0",
"wcwidth" : "0.2.13",
"websockets" : "15.0.1",
"wrapt" : "1.17.3",
"wsproto" : "1.2.0",
"zipp" : "3.23.0"
},
"config": {
"schema_version": "4.0.2",
"legacy_models_yaml_path": null,
"host": "127.0.0.1",
"port": 9090,
"allow_origins": [],
"allow_credentials": true,
"allow_methods": [""],
"allow_headers": [""],
"ssl_certfile": null,
"ssl_keyfile": null,
"log_tokenization": false,
"patchmatch": true,
"models_dir": "models",
"convert_cache_dir": "models/.convert_cache",
"download_cache_dir": "models/.download_cache",
"legacy_conf_dir": "configs",
"db_dir": "databases",
"outputs_dir": "outputs",
"custom_nodes_dir": "nodes",
"style_presets_dir": "style_presets",
"workflow_thumbnails_dir": "workflow_thumbnails",
"log_handlers": ["console"],
"log_format": "color",
"log_level": "info",
"log_sql": false,
"log_level_network": "warning",
"use_memory_db": false,
"dev_reload": false,
"profile_graphs": false,
"profile_prefix": null,
"profiles_dir": "profiles",
"max_cache_ram_gb": null,
"max_cache_vram_gb": null,
"log_memory_usage": false,
"device_working_mem_gb": 3,
"enable_partial_loading": true,
"keep_ram_copy_of_weights": true,
"ram": null,
"vram": null,
"lazy_offload": true,
"pytorch_cuda_alloc_conf": null,
"device": "auto",
"precision": "auto",
"sequential_guidance": false,
"attention_type": "auto",
"attention_slice_size": "auto",
"force_tiled_decode": false,
"pil_compress_level": 1,
"max_queue_size": 10000,
"clear_queue_on_startup": false,
"allow_nodes": null,
"deny_nodes": null,
"node_cache_size": 512,
"hashing_algorithm": "blake3_single",
"remote_api_tokens": null,
"scan_models_on_startup": false,
"unsafe_disable_picklescan": false
},
"set_config_fields": ["legacy_models_yaml_path", "enable_partial_loading"]
}

What happened

I enabled the NSFW checker and NSFW images are still produced and shown. This is the error:

[2025-09-20 23:21:05,520]::[InvokeAI]::WARNING --> Could not load NSFW checker: Unrecognized model in /Users/me/invoke-root/models/core/convert/stable-diffusion-safety-checker. Should have a `model_type` key in its config.json, or contain one of the following strings in its name: aimv2, aimv2_vision_model, albert, align, altclip, apertus, arcee, aria, aria_text, audio-spectrogram-transformer, autoformer, aya_vision, bamba, bark, bart, beit, bert, bert-generation, big_bird, bigbird_pegasus, biogpt, bit, bitnet, blenderbot, blenderbot-small, blip, blip-2, blip_2_qformer, bloom, bridgetower, bros, camembert, canine, chameleon, chinese_clip, chinese_clip_vision_model, clap, clip, clip_text_model, clip_vision_model, clipseg, clvp, code_llama, codegen, cohere, cohere2, cohere2_vision, colpali, colqwen2, conditional_detr, convbert, convnext, convnextv2, cpmant, csm, ctrl, cvt, d_fine, dab-detr, dac, data2vec-audio, data2vec-text, data2vec-vision, dbrx, deberta, deberta-v2, decision_transformer, deepseek_v2, deepseek_v3, deepseek_vl, deepseek_vl_hybrid, deformable_detr, deit, depth_anything, depth_pro, deta, detr, dia, diffllama, dinat, dinov2, dinov2_with_registers, dinov3_convnext, dinov3_vit, distilbert, doge, donut-swin, dots1, dpr, dpt, efficientformer, efficientloftr, efficientnet, electra, emu3, encodec, encoder-decoder, eomt, ernie, ernie4_5, ernie4_5_moe, ernie_m, esm, evolla, exaone4, falcon, falcon_h1, falcon_mamba, fastspeech2_conformer, fastspeech2_conformer_with_hifigan, flaubert, flava, florence2, fnet, focalnet, fsmt, funnel, fuyu, gemma, gemma2, gemma3, gemma3_text, gemma3n, gemma3n_audio, gemma3n_text, gemma3n_vision, git, glm, glm4, glm4_moe, glm4v, glm4v_moe, glm4v_moe_text, glm4v_text, glpn, got_ocr2, gpt-sw3, gpt2, gpt_bigcode, gpt_neo, gpt_neox, gpt_neox_japanese, gpt_oss, gptj, gptsan-japanese, granite, granite_speech, granitemoe, granitemoehybrid, granitemoeshared, granitevision, graphormer, grounding-dino, groupvit, helium, hgnet_v2, hiera, hubert, hunyuan_v1_dense, hunyuan_v1_moe, ibert, idefics, idefics2, idefics3, idefics3_vision, ijepa, imagegpt, informer, instructblip, instructblipvideo, internvl, internvl_vision, jamba, janus, jetmoe, jukebox, kosmos-2, kosmos-2.5, kyutai_speech_to_text, layoutlm, layoutlmv2, layoutlmv3, led, levit, lfm2, lightglue, lilt, llama, llama4, llama4_text, llava, llava_next, llava_next_video, llava_onevision, longformer, longt5, luke, lxmert, m2m_100, mamba, mamba2, marian, markuplm, mask2former, maskformer, maskformer-swin, mbart, mctct, mega, megatron-bert, metaclip_2, mgp-str, mimi, minimax, mistral, mistral3, mixtral, mlcd, mllama, mm-grounding-dino, mobilebert, mobilenet_v1, mobilenet_v2, mobilevit, mobilevitv2, modernbert, modernbert-decoder, moonshine, moshi, mpnet, mpt, mra, mt5, musicgen, musicgen_melody, mvp, nat, nemotron, nezha, nllb-moe, nougat, nystromformer, olmo, olmo2, olmoe, omdet-turbo, oneformer, open-llama, openai-gpt, opt, ovis2, owlv2, owlvit, paligemma, patchtsmixer, patchtst, pegasus, pegasus_x, perceiver, perception_encoder, perception_lm, persimmon, phi, phi3, phi4_multimodal, phimoe, pix2struct, pixtral, plbart, poolformer, pop2piano, prompt_depth_anything, prophetnet, pvt, pvt_v2, qdqbert, qwen2, qwen2_5_omni, qwen2_5_vl, qwen2_5_vl_text, qwen2_audio, qwen2_audio_encoder, qwen2_moe, qwen2_vl, qwen2_vl_text, qwen3, qwen3_moe, rag, realm, recurrent_gemma, reformer, regnet, rembert, resnet, retribert, roberta, roberta-prelayernorm, roc_bert, roformer, rt_detr, rt_detr_resnet, rt_detr_v2, rwkv, sam, sam2, sam2_hiera_det_model, sam2_video, sam2_vision_model, sam_hq, sam_hq_vision_model, sam_vision_model, seamless_m4t, seamless_m4t_v2, seed_oss, segformer, seggpt, sew, sew-d, shieldgemma2, siglip, siglip2, siglip_vision_model, smollm3, smolvlm, smolvlm_vision, speech-encoder-decoder, speech_to_text, speech_to_text_2, speecht5, splinter, squeezebert, stablelm, starcoder2, superglue, superpoint, swiftformer, swin, swin2sr, swinv2, switch_transformers, t5, t5gemma, table-transformer, tapas, textnet, time_series_transformer, timesfm, timesformer, timm_backbone, timm_wrapper, trajectory_transformer, transfo-xl, trocr, tvlt, tvp, udop, umt5, unispeech, unispeech-sat, univnet, upernet, van, video_llava, videomae, vilt, vipllava, vision-encoder-decoder, vision-text-dual-encoder, visual_bert, vit, vit_hybrid, vit_mae, vit_msn, vitdet, vitmatte, vitpose, vitpose_backbone, vits, vivit, vjepa2, voxtral, voxtral_encoder, wav2vec2, wav2vec2-bert, wav2vec2-conformer, wavlm, whisper, xclip, xcodec, xglm, xlm, xlm-prophetnet, xlm-roberta, xlm-roberta-xl, xlnet, xlstm, xmod, yolos, yoso, zamba, zamba2, zoedepth

What you expected to happen

I think the Invoke NSFW checker is supposed to blur the images.

How to reproduce the problem

Enable NSFW checker and use a NSFW prompt.

Additional context

No response

Discord username

magnusviri

[psychedelicious]

Not sure when this broke. Reading the NSFW checker code, I see it doesn't have the right setup to load the model. We'll need to refactor the NSFW checker backend and node to fix it. It needs to be updated to use the InvocationContext's load_remote_model API.

That said, I think we should either:

1. Remove the NSFW checker (and invisible watermark) entirely.
2. Refactor these features so that they are backend-only features that operate automatically.

To elaborate on 2:

Today, NSFW checking (and invisible watermarking) is accomplished by adding nodes into our Canvas/Generate tab graphs. When a NSFW image is detected by these nodes, it is blurred but the graph continues executing. Invisible watermark simply adds the watermark.

There are problems with this approach:

• It doesn't work on the Workflow Editor.
• Only the final output is checked. Images created earlier in the graph could be NSFW but we don't check them.
• If NSFW is detected very early on in the graph and we blur the image, there's no reason to continue the graph because we already know this output is not wanted. But the graph does continue to execute, without "knowing" it's had a NSFW detection. It just operates on the blurred iamge.

The proposed refactor would instead run NSFW checking and watermarking as images are saved:

• If NSFW checking is enabled, before we save any image, we check for NSFW. If NSFW is detected, raise an error and stop graph execution immediately.
• If watermarking is enabled, before we save any image, we apply the watermark.

We would do the check and watermark in the InvocationContext.images.save() method. Opting in to these features would be done by editing invokeai.yaml, not via UI.

[psychedelicious]

I reviewed the NSFW checker again and realized it does have everything it needs to load the model - it's just doing it in a way that doesn't use the model manager for memory management. While that is not ideal, it does work and successfully blurs NSFW images for me.

Maybe your NSFW checker model is corrupted? Try deleting its folder, then restart Invoke and try again. It's located in <invoke_models_dir>/core/convert/stable-diffusion-safety-checker.

I think the proposed refactor is still a good idea, but not necessary at this time

[magnusviri]

Ok, that was silly. I deleted the stable-diffusion-safety-checker folder and tried it and it worked. I have no idea how it broke. The only file in the broken version was preprocessor_config.json.

It would be nice if the NSFW filter also blurred the preview image until it passes the checker. And maybe more blur lol. I don't think I can change this issue to a feature request. Should I close this and reopen it as a feature request?