A personal fork of lllyasviel/Fooocus v2.5.5 with quality-of-life features: Save Preset, CivitAI Model Settings, LoRA trigger words, Embeddings panel, Wildcards editor, Vary-with-aspect-ratio, Custom Resolution, Asset Browser, Restart UI button.
Apps by @supersoniquestudio
17 totalQwen-Image studio + instruction image editing (Qwen-Image-Edit-2509), Fooocus-style, 100% local: txt2img, ESRGAN+Qwen refine upscale, LoRA, styles, Describe/Improve & Vision Mix (Ollama), Remove BG, Reframe/outpaint, Face Swap, and an Edit tab (image + prompt). Fork of crispz-studio. https://github.com/mikecastrodemaria/crispz-qwen-edit
Krea 2 txt2img studio (Fooocus-style, 100% local): txt2img, pure-ESRGAN upscale, single-file/Civitai models, LoRA, styles, Describe/Improve & Vision Mix (Ollama), Remove BG, Face Swap. https://github.com/mikecastrodemaria/crispz-krea2
Standalone local hi-res fix: enlarge images with Real-ESRGAN and reinject clean detail with Z-Image Turbo img2img. 100% local — no ComfyUI, no SwarmUI, no cloud.
Z-Image txt2img + upscaler/detailer studio (Fooocus-style, 100% local): txt2img, ESRGAN+Z-Image refine upscale, single-file/Civitai models, LoRA, styles, Describe/Improve & Vision Mix (Ollama), Remove BG, Reframe/outpaint, Face Swap. https://github.com/mikecastrodemaria/crispz-studio
FLUX.1 Krea [dev] txt2img studio (Fooocus-style, 100% local): high-aesthetic text-to-image, ESRGAN+Flux refine upscale, single-file/Civitai Flux models, LoRA, styles, Describe/Improve & Vision Mix (Ollama), Remove BG, Reframe/outpaint, Face Swap. Fork of crispz-studio. https://github.com/mikecastrodemaria/crispz-krea
Pre-mastering & audio enhancement for AI-generated music. 12-stage processing chain with platform presets (Suno, Udio), before/after spectrogram, and broadcast-ready LUFS normalization.
Pre-mastering & audio enhancement for AI-generated music. 12-stage processing chain with platform presets (Suno, Udio), before/after spectrogram, and broadcast-ready LUFS normalization.
FLUX.2 Klein 4B studio (Apache 2.0, 4 steps), Fooocus-style, 100% local: txt2img, multi-reference editing in the SAME pipeline (up to 4 refs, no second model), inpaint/outpaint, ESRGAN+refine upscale, LoRA, styles, Describe/Improve & Vision Mix (Ollama), Remove BG, Face Swap. ~15 GB VRAM for everything. Note: klein is distilled, so negative prompts and CFG have no effect. Fork of crispz-qwen-edit. https://github.com/mikecastrodemaria/crispz-klein
A web interface for the Moondream3 vision-language model featuring image captioning, visual question answering, object detection, and object pointing.
FLUX.3 Image Klein studio (5.6B, 16 steps, FLUX Kommunity License), Fooocus-style, 100% local: txt2img, multi-reference editing in the SAME pipeline (<ref_image_0>, <ref_image_1>..., no second model), inpaint/outpaint, ESRGAN+refine upscale, LoRA, styles, Describe/Improve & Vision Mix (Ollama), Remove BG, Face Swap. 18.8 GB of VRAM in bf16 for everything (fp8r / nvfp4 builds for smaller cards). The repos are gated on Hugging Face: accept the licence and set a token. Fork of crispz-klein. https://github.com/mikecastrodemaria/crispz-flux3
Batch image captioning with Ollama vision models: composed prompts (type × length × options), image preview, editable captions, skip/overwrite/append, live progress. Web UI (NiceGUI) or Gradio UI on the same core. https://github.com/mikecastrodemaria/Captionz
Customizing Realistic Human Photos via Stacked ID Embedding with SDXL model switching support. Load any local SDXL .safetensors checkpoint directly from the UI.
MOSS-Image studio (5B open weights, Apache 2.0: commercial use allowed), Fooocus-style, 100% local: the text-in-image and design studio of the crispz family (posters, logos, packaging, dense typography). Layout JSON with boxes drawn on a canvas, local magic prompt and Describe (Ollama or the model's own Qwen3.5-9B encoder), 9 design presets, emulated img2img / inpaint / outpaint, ESRGAN upscale, LoRAs in every published layout, editing with references (Moss-Image-Edit LoRA), local vectorization (SVG) and Remove BG, the MOSS-Image Character Catalog, hand-overs to crispz-flux3 and crispz-ideogram. NVIDIA GPU with 24 GB or more advised (27 GB of bf16 weights, encoder-only offload). Public weights: no licence to accept, no token. https://github.com/mikecastrodemaria/crispz-moss
Local, engine-agnostic comic book workshop: script -> flatplan -> pages -> balloons -> PDF. Draws through the crispz family apps (crispz-studio, crispz-klein, crispz-qwen-edit...) over the CLI protocol; panel shapes, break-the-frame cutouts, frameless elements, Visual Bible, storyboard, .md bundles. https://github.com/mikecastrodemaria/comics2crispz
Ideogram 4.0 studio (open weights, NON-COMMERCIAL licence), Fooocus-style, 100% local: the text-in-image and design studio of the crispz family (posters, logos, packaging, dense typography). Structured JSON captions with boxes drawn on a canvas, local magic prompt and Describe (Qwen3-VL-8B), 9 design presets, emulated img2img / inpaint / outpaint, ESRGAN upscale, local vectorization (SVG) and Remove BG, editing handed to crispz-flux3. NVIDIA GPU required (nf4 weights, bitsandbytes). The repo is gated on Hugging Face: accept the licence and set a token. https://github.com/mikecastrodemaria/crispz-ideogram
Short videos up to 4K from a few key images made by the crispz family engines (crispz-moss + Turbo LoRA by default), FILM frame interpolation in between, ESRGAN to 4K, ffmpeg exports. Web page with keys strip, img2img chain, edit, inpaint, middle keys, distance guard, and import of the A1111 Animator settings (settings.txt, keyframes.csv). No video model. https://github.com/mikecastrodemaria/crispz-motion
