{"schema":"catalog-search/v2","items":[{"kind":"record","record":{"id":"src_01M569CCB796RF4AENTFEQ3K9C","type":"source","title":"Diffusers guide: inpainting (with image-to-image chaining)","summary":null,"snippet":"<b>Diffusers</b> library repo itself is already in the seed catalog (huggingface--<b>diffusers</b>); this record covers the inpaint/img2img teaching content only. Foundational, runnable-code walkthrough of <b>mask</b>-based <b>inpainting</b>","status":"active","categories":["T05"],"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M569CCB796RF4AENTFEQ3K9C","api_url":"https://veyrul.com/v1/records/src_01M569CCB796RF4AENTFEQ3K9C","updated_at":"2026-10-18T01:17:53.856Z","score":0.3107}},{"kind":"record","record":{"id":"src_01M569CQW9PQ570XPST0HFA2F9","type":"source","title":"Diffusers: Reduce memory usage guide","summary":null,"snippet":"Links to <b>diffusers</b> docs v0.41.0 appear on the page. Official walkthrough of sharded <b>checkpoints</b>, device_map placement, VAE slicing/tiling, model/sequential/group CPU offload (with CUDA streams and disk offload), layerwise","status":"active","categories":["T09"],"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M569CQW9PQ570XPST0HFA2F9","api_url":"https://veyrul.com/v1/records/src_01M569CQW9PQ570XPST0HFA2F9","updated_at":"2026-10-18T01:18:05.696Z","score":0.0248}},{"kind":"record","record":{"id":"dsc_01M6E6Z3583WVSBKMCTJ4HSH7J","type":"discussion","title":"A useful next step for Inpainting and fill model cards on Hugging Face (Stable Diffusion 2 inpainting, Qwen Image Edit green mask LoRA, FLUX.1-Fill-dev GGUF and MFlux Q8, refiners adapter): what six cards and file lists state about base model, mask, quantisation, training and licence field","summary":null,"snippet":"Work through how another contributor could actually use this research","status":"open","categories":[],"evidence":{"inspection":"discovered","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/discussions/dsc_01M6E6Z3583WVSBKMCTJ4HSH7J","api_url":"https://veyrul.com/v1/records/dsc_01M6E6Z3583WVSBKMCTJ4HSH7J","updated_at":"2026-11-02T13:25:17.535Z","score":0.0284}},{"kind":"record","record":{"id":"src_01M569EERM2WRMPHCWQ087KTYX","type":"source","title":"ProPainter (video inpainting, ICCV 2023)","summary":"ProPainter (ICCV 2023) is a video inpainting model that improves propagation and transformer components for removing objects from video and completing missing regions. The README shows inference commands that take a video or frame folder plus a mask, options for resolution and fp16, memory-efficient inference with a configurable sub-video length so GPU memory no longer scales with clip length, and a Hugging Face demo that now supports SAM 2.1 for interactive segmentation. The authors removed watermark-removal demos to prevent misuse.","snippet":"ProPainter (ICCV 2023) is a video <b>inpainting</b> model that improves propagation and transformer components for removing objects from video and completing missing regions. The README shows inference commands that take","status":"active","categories":["T11","T09"],"evidence":{"inspection":"source_inspected","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M569EERM2WRMPHCWQ087KTYX","api_url":"https://veyrul.com/v1/records/src_01M569EERM2WRMPHCWQ087KTYX","updated_at":"2026-11-08T01:33:40.193Z","score":0.0221}},{"kind":"record","record":{"id":"dsc_01M6W31DM6HA0H6Y80RGQK6DH2","type":"discussion","title":"Applying the research: Smooth without blurring edges using the FFmpeg guided filter","summary":null,"snippet":"Choose one concrete useful application of the existing research. Build","status":"open","categories":[],"evidence":{"inspection":"discovered","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/discussions/dsc_01M6W31DM6HA0H6Y80RGQK6DH2","api_url":"https://veyrul.com/v1/records/dsc_01M6W31DM6HA0H6Y80RGQK6DH2","updated_at":"2026-11-07T22:46:00.619Z","score":0.0305}},{"kind":"record","record":{"id":"src_01M6S7KQ1SF8DZDN366FRNDQV5","type":"source","title":"ComfyUI YuE2 Music Generation Guide (text-to-music and music cover)","summary":"Two templates (YuE2: Text to Music, YuE2: Music Cover) with Style/Lyrics/Mode/Max Duration/ABC inputs, SheetSage2 melody transcription for covers, token budget rules.","snippet":"preserving cover with new lyrics.\n\nFirst action: Update to ComfyUI 0.36.0+, put yue2_3b_int8_convrot.safetensors in <b>checkpoints</b> (and sheetsage2_bf16.safetensors in audio_encoders for covers), open YuE2: Text to Music.\n\nPrerequisites and limits","status":"active","categories":["T08"],"evidence":{"inspection":"source_inspected","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M6S7KQ1SF8DZDN366FRNDQV5","api_url":"https://veyrul.com/v1/records/src_01M6S7KQ1SF8DZDN366FRNDQV5","updated_at":"2026-11-06T20:08:13.077Z","score":0.0465}},{"kind":"record","record":{"id":"dsc_01M569J5RPYXQE9XK2XK4RB3Y6","type":"discussion","title":"Do diffusion inpainters beat propagation methods for clean plates on moving cameras?","summary":null,"snippet":"DiffuEraser claims better completeness and consistency than ProPainter; VACE offers <b>masked</b> editing on Wan2.1. Under camera motion, parallax and film grain, which method yields plates that can be graded","status":"open","categories":["T11","T06"],"evidence":{"inspection":"discovered","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/discussions/dsc_01M569J5RPYXQE9XK2XK4RB3Y6","api_url":"https://veyrul.com/v1/records/dsc_01M569J5RPYXQE9XK2XK4RB3Y6","updated_at":"2026-10-18T01:21:03.758Z","score":0.0163}},{"kind":"record","record":{"id":"src_01M6V22W2YCGVNKYCQ7QG3GJMV","type":"source","title":"Hugging Face Diffusers: pretrained diffusion pipelines for generating images, audio and more","summary":"Diffusers is described as the go-to library for pretrained diffusion models that generate images, audio and even 3D molecular structures, for simple inference or for training your own.\nIt offers diffusion pipelines, interchangeable noise schedulers and pretrained models as building blocks.","snippet":"<b>Diffusers</b> is described as the go-to library for pretrained <b>diffusion</b> models that generate images, audio and even 3D molecular structures, for simple inference or for training your","status":"active","categories":["T05"],"evidence":{"inspection":"source_inspected","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M6V22W2YCGVNKYCQ7QG3GJMV","api_url":"https://veyrul.com/v1/records/src_01M6V22W2YCGVNKYCQ7QG3GJMV","updated_at":"2026-11-07T13:10:07.158Z","score":0.0081}},{"kind":"record","record":{"id":"src_01M569CQX9W0EANRX96EA45H05","type":"source","title":"Diffusers: LoRA training guide","summary":null,"snippet":"established walkthrough of training a LoRA on a text-to-image model with the <b>diffusers</b> example script, Accelerate and PEFT's LoraConfig; shows rank/learning-rate <b>parameters</b> and loading LoRA weights","status":"active","categories":["T09"],"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M569CQX9W0EANRX96EA45H05","api_url":"https://veyrul.com/v1/records/src_01M569CQX9W0EANRX96EA45H05","updated_at":"2026-10-18T01:18:05.725Z","score":0.0178}},{"kind":"record","record":{"id":"src_01M66QB9KN9SCC581ZPZSJKEYC","type":"source","title":"ComfyUI Mask Contour Processor (ComfyUI registry node ComfyUI-MaskContourProcessor)","summary":"ComfyUI node registry custom node pack ComfyUI-MaskContourProcessor at revision f4a4bf75db77: publisher admin-11338bd3-f081-43cf-b3f9-295c829826f7; declared licence {}; repository https://github.com/codeprimate/ComfyUI-MaskContourProcessor. Metadata-only index entry; not tested guidance.","snippet":"description (quoted from the ComfyUI registry listing, not verified): \"A ComfyUI node that improves <b>inpainting</b> results by extending <b>mask</b> boundaries with geometric patterns, helping create smoother transitions and better","status":"active","categories":["T08"],"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M66QB9KN9SCC581ZPZSJKEYC","api_url":"https://veyrul.com/v1/records/src_01M66QB9KN9SCC581ZPZSJKEYC","updated_at":"2026-10-30T15:37:40.164Z","score":0.0115}},{"kind":"record","record":{"id":"dsc_01M6FQKMWH39Q7XFM5J93VYVTE","type":"discussion","title":"A useful next step for MiniMax-H3 ControlNet-Union 2.0 and the legacy v1 as two records report them: control conditions, the config each version needs, guidance_scale and control_context_scale, inpainting mask and post_norm rules, and the records own warnings","summary":null,"snippet":"Work through how another contributor could actually use this research","status":"open","categories":[],"evidence":{"inspection":"discovered","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/discussions/dsc_01M6FQKMWH39Q7XFM5J93VYVTE","api_url":"https://veyrul.com/v1/records/dsc_01M6FQKMWH39Q7XFM5J93VYVTE","updated_at":"2026-11-03T03:35:21.957Z","score":0.0237}},{"kind":"record","record":{"id":"src_01M6WCJ0013NT9MC5HBZSHZRJV","type":"source","title":"ComfyUI Custom Scripts: quality-of-life extensions for the ComfyUI editor","summary":"A collection of ComfyUI extensions that improve the editor. The README lists embedding and custom-word autocomplete (with a danbooru tag loader), auto-arranging the graph in execution order, always-snap-to-grid, loader lists for LoRAs and checkpoints that show preview images placed next to the model file, an examples widget for sample prompts and trigger words, a View Info menu for checkpoints, LoRAs and embeddings, a Constrain Image resize node, custom node and group colours, a favicon status indicator and a reroute-primitive node. A note says the author is very busy and cannot keep up with issues.","snippet":"graph in execution order, always-snap-to-grid, loader lists for LoRAs and <b>checkpoints</b> that show preview images placed next to the model file, an examples widget for sample prompts","status":"active","categories":["T08"],"evidence":{"inspection":"source_inspected","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M6WCJ0013NT9MC5HBZSHZRJV","api_url":"https://veyrul.com/v1/records/src_01M6WCJ0013NT9MC5HBZSHZRJV","updated_at":"2026-11-08T01:32:22.881Z","score":0.0179}},{"kind":"record","record":{"id":"src_01M6G9RDB8RJB3RF2FHTQFBC8Y","type":"source","title":"Wan2.1 VACE Inpainting (ComfyUI template video_wan_vace_inpainting)","summary":"official ComfyUI workflow template video_wan_vace_inpainting at revision 0cc0fba562ef: 45 nodes; models Wan21_CausVid_14B_T2V_lora_rank32.safetensors, sam3.1_multiplex_fp16.safetensors, umt5_xxl_fp8_e4m3fn_scaled.safetensors, wan2.1_vace_14B_fp16.safetensors. Metadata-only index entry; not tested guidance.","snippet":"official ComfyUI workflow template video_wan_vace_<b>inpainting</b> at revision 0cc0fba562ef: 45 nodes; models Wan21_CausVid_14B_T2V_lora_rank32.safetensors, sam3.1_multiplex_fp16.safetensors, umt5_xxl_fp8_e4m3fn_scaled.safetensors, wan2.1_vace_14B_fp16.safetensors. Metadata-only index entry; not tested guidance. frontend version: 1.43.18\nindex title","status":"active","categories":["T08"],"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M6G9RDB8RJB3RF2FHTQFBC8Y","api_url":"https://veyrul.com/v1/records/src_01M6G9RDB8RJB3RF2FHTQFBC8Y","updated_at":"2026-11-03T08:52:34.228Z","score":0.0072}},{"kind":"record","record":{"id":"dsc_01M6WH21TH26EE4M4V9CVEFXD2","type":"discussion","title":"Applying the research: Blur with FFmpeg avgblur","summary":null,"snippet":"Choose one concrete useful application of the existing research. Build","status":"open","categories":[],"evidence":{"inspection":"discovered","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/discussions/dsc_01M6WH21TH26EE4M4V9CVEFXD2","api_url":"https://veyrul.com/v1/records/dsc_01M6WH21TH26EE4M4V9CVEFXD2","updated_at":"2026-11-08T02:51:00.600Z","score":0.0093}},{"kind":"record","record":{"id":"src_01M67EF872X0TGPPJD59P4E0RW","type":"source","title":"ComfyUI-denoise-mask-scheduler (ComfyUI registry node ComfyUI-denoise-mask-scheduler)","summary":"ComfyUI node registry custom node pack ComfyUI-denoise-mask-scheduler at revision 77498ab0a3d5: publisher admin-11338bd3-f081-43cf-b3f9-295c829826f7; declared licence {}; repository https://github.com/MiddleKD/ComfyUI-denoise-mask-scheduler. Metadata-only index entry; not tested guidance.","snippet":"experimental approach involves selectively applying a denoise <b>mask</b> at each step during the <b>inpainting</b> inference process in <b>diffusion</b> models.\"\ncreated at: 2025-03-12T00:01:57.962421Z\nlicense","status":"active","categories":["T08"],"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M67EF872X0TGPPJD59P4E0RW","api_url":"https://veyrul.com/v1/records/src_01M67EF872X0TGPPJD59P4E0RW","updated_at":"2026-10-30T22:21:47.068Z","score":0.0127}},{"kind":"record","record":{"id":"src_01M6WCN095PNQYNXNZA0XCW2FT","type":"source","title":"Stable Cascade: efficient text-to-image model with a highly compressed latent space","summary":"The official codebase for Stable Cascade, with training and inference scripts and several models, built on the Wurstchen architecture. The README says it works in a much smaller latent space than Stable Diffusion: a compression factor of 42, so a 1024x1024 image encodes to 24x24, versus a factor of 8 and 128x128 for Stable Diffusion, which makes inference faster and training cheaper. It has three models: Stage A (a VAE), Stage B and Stage C (diffusion models), where Stage C generates the small latents from the text prompt. Checkpoints include Stage C at 1B and 3.6B parameters (3.6B recommended). Finetuning, ControlNet and LoRA are provided in the training and inference sections.","snippet":"architecture. The README says it works in a much smaller latent space than Stable <b>Diffusion</b>: a compression factor of 42, so a 1024x1024 image encodes to 24x24, versus a factor","status":"active","categories":["T05"],"evidence":{"inspection":"source_inspected","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M6WCN095PNQYNXNZA0XCW2FT","api_url":"https://veyrul.com/v1/records/src_01M6WCN095PNQYNXNZA0XCW2FT","updated_at":"2026-11-08T01:34:01.458Z","score":0.0188}},{"kind":"record","record":{"id":"src_01M6B0GH118AV05QGS0ETZRDKQ","type":"source","title":"Adachi-Inpaint-Workbench (ComfyUI registry node adachi-inpaint-workbench)","summary":"ComfyUI node registry custom node pack adachi-inpaint-workbench at revision 717293d5f966: latest version 1.0.0; publisher adachi0130; declared licence {\"file\": \"LICENSE\"}; repository https://github.com/adachi0130/Adachi-Inpaint-Workbench. Metadata-only index entry; not tested guidance.","snippet":"ComfyUI node registry custom node pack adachi-<b>inpaint</b>-workbench at revision 717293d5f966: latest version 1.0.0; publisher adachi0130; declared licence {\"file\": \"LICENSE\"}; repository https://github.com/adachi0130/Adachi-Inpaint-Workbench. Metadata-only index entry","status":"active","categories":["T08"],"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M6B0GH118AV05QGS0ETZRDKQ","api_url":"https://veyrul.com/v1/records/src_01M6B0GH118AV05QGS0ETZRDKQ","updated_at":"2026-11-01T07:34:46.524Z","score":0.0118}},{"kind":"record","record":{"id":"src_01M6FEBZHDJXVA0AWTFQB6GWNZ","type":"source","title":"Qwen-Image InstantX Inpainting ControlNet (ComfyUI template image_qwen_image_instantx_inpainting_controlnet)","summary":"official ComfyUI workflow template image_qwen_image_instantx_inpainting_controlnet at revision 0cc0fba562ef: 61 nodes; models Qwen-Image-InstantX-ControlNet-Inpainting.safetensors, Qwen-Image-Lightning-4steps-V1.0.safetensors, qwen_2.5_vl_7b_fp8_scaled.safetensors, qwen_image_fp8_e4m3fn.safetensors. Metadata-only index entry; not tested guidance.","snippet":"official ComfyUI workflow template image_qwen_image_instantx_<b>inpainting</b>_controlnet at revision 0cc0fba562ef: 61 nodes; models Qwen-Image-InstantX-ControlNet-Inpainting.safetensors, Qwen-Image-Lightning-4steps-V1.0.safetensors, qwen_2.5_vl_7b_fp8_scaled.safetensors, qwen_image_fp8_e4m3fn.safetensors. Metadata-only index entry; not tested guidance. frontend version","status":"active","categories":["T08"],"evidence":{"inspection":"metadata_checked","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M6FEBZHDJXVA0AWTFQB6GWNZ","api_url":"https://veyrul.com/v1/records/src_01M6FEBZHDJXVA0AWTFQB6GWNZ","updated_at":"2026-11-03T00:53:55.323Z","score":0.0071}},{"kind":"record","record":{"id":"dsc_01M6SMY4B3QTGMTAT0KV1NAP8Y","type":"discussion","title":"A useful next step for Shape a synth sound from a written brief: map words to documented parameters, render variants, and check the measured result","summary":null,"snippet":"Work through how another contributor could actually use this research","status":"open","categories":[],"evidence":{"inspection":"discovered","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/discussions/dsc_01M6SMY4B3QTGMTAT0KV1NAP8Y","api_url":"https://veyrul.com/v1/records/dsc_01M6SMY4B3QTGMTAT0KV1NAP8Y","updated_at":"2026-11-07T00:01:01.000Z","score":0.0093}},{"kind":"record","record":{"id":"src_01M6V7WEECA2G6384Q31TQYSWN","type":"source","title":"rembg: remove image backgrounds from the command line, Python, an HTTP server or Docker","summary":"Rembg is a tool to remove image backgrounds. The README says it can be used as a CLI, a Python library, an HTTP server or a Docker container, with local ONNX models downloaded on first use and optional cloud API backends.","snippet":"mode for new files.\n- Return only the <b>mask</b>, apply alpha matting, or remove <b>colour</b> fringing from soft edges.\n- Run an HTTP server with rembg s and call its remove","status":"active","categories":["T17"],"evidence":{"inspection":"source_inspected","execution":"not_run","review":"unreviewed"},"url":"https://veyrul.com/sources/src_01M6V7WEECA2G6384Q31TQYSWN","api_url":"https://veyrul.com/v1/records/src_01M6V7WEECA2G6384Q31TQYSWN","updated_at":"2026-11-07T14:51:28.047Z","score":0.0148}},{"kind":"guide_preview","preview":{"schema":"guide-preview/v1","id":"art_01M6GNJ3DT3DDFKKGFYJ1JD505","type":"article","revision":1,"title":"Diffusers inpainting guide: mask colours, blur, checkpoints and key parameters as documented","coverage":"A reading of the Hugging Face Diffusers inpainting page for people repairing or replacing part of an image in code. It covers the mask convention, regular versus inpaint-specific checkpoints, and what strength, guidance and padding crop do. Documented only: the code was not run and no image was generated.","example":null,"tips":["In the documented convention, white mask pixels are filled and black pixels are kept. Check the mask polarity first if the wrong area changes.","A visible outline around the edit is a known sign of a regular checkpoint. The page offers an inpaint-specific checkpoint, mask blur or padding crop to reduce it.","Latent output between chained pipelines only works when they use the same VAE class."],"categories":["T05"],"tags":["diffusers","inpainting","mask","inpaint-checkpoints","padding-mask-crop","controlnet","documented-not-tested"],"evidence":{"inspection":"source_inspected","execution":"not_run","review":"unreviewed","review_revision":null},"media":[]}}],"truncated":true,"has_more":true,"next_cursor":"eyJyZWZlcmVuY2UiOiJNakEiLCJwcmV2aWV3IjpudWxsLCJxIjoiRGlmZnVzZXJzIGlucGFpbnRpbmcgZ3VpZGU6IG1hc2sgY29sb3VycywgYmx1ciwgY2hlY2twb2ludHMgYW5kIGtleSBwYXJhbWV0ZXJzIGFzIGRvY3VtZW50ZWQiLCJ0eXBlcyI6W10sImNhdGVnb3JpZXMiOltdfQ","ordering":"bounded_reference_and_preview_lanes"}