{"id":47279835,"url":"https://github.com/modl-org/modl","last_synced_at":"2026-04-02T00:10:42.687Z","repository":{"id":340006846,"uuid":"1164149231","full_name":"modl-org/modl","owner":"modl-org","description":"Local-first AI image generation toolkit. Pull models, train LoRAs, generate images. One CLI, no glue code.","archived":false,"fork":false,"pushed_at":"2026-03-25T22:17:37.000Z","size":8243,"stargazers_count":9,"open_issues_count":16,"forks_count":2,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-03-25T23:44:56.742Z","etag":null,"topics":["ai","cli","diffusion","flux","generative-ai","image-generation","local-first","lora","rust","self-hosted","stable-diffusion"],"latest_commit_sha":null,"homepage":"https://modl.run","language":"Rust","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"agpl-3.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/modl-org.png","metadata":{"files":{"readme":"README.md","changelog":"CHANGELOG.md","contributing":"CONTRIBUTING.md","funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-02-22T18:00:09.000Z","updated_at":"2026-03-25T22:17:41.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/modl-org/modl","commit_stats":null,"previous_names":["pedropaf/mods","modshq-org/mods","modl-org/modl"],"tags_count":18,"template":false,"template_full_name":null,"purl":"pkg:github/modl-org/modl","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/modl-org%2Fmodl","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/modl-org%2Fmodl/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/modl-org%2Fmodl/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/modl-org%2Fmodl/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/modl-org","download_url":"https://codeload.github.com/modl-org/modl/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/modl-org%2Fmodl/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31293286,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-01T21:15:39.731Z","status":"ssl_error","status_checked_at":"2026-04-01T21:15:34.046Z","response_time":53,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai","cli","diffusion","flux","generative-ai","image-generation","local-first","lora","rust","self-hosted","stable-diffusion"],"created_at":"2026-03-15T22:51:23.822Z","updated_at":"2026-04-02T00:10:42.674Z","avatar_url":"https://github.com/modl-org.png","language":"Rust","funding_links":[],"categories":[],"sub_categories":[],"readme":"# modl\n\n**Train LoRAs and generate images on your own GPU.** Web UI + CLI. Managed runtime. It just works.\n\n```bash\ncurl -fsSL https://modl.run/install.sh | sh\nmodl pull z-image-turbo\nmodl generate \"a cat on mars\"\n```\n\n**[Website](https://modl.run)** · **[Docs](https://modl.run/docs)** · **[Guides](https://modl.run/guides)** · **[Model Registry](https://github.com/modl-org/modl-registry)** · **[Changelog](CHANGELOG.md)**\n\n---\n\n## Why modl?\n\n**No glue code.** One binary handles model downloads, dependency resolution, image generation, LoRA training, and output management. No separate tools to install, no configs to write.\n\n**Smart model management.** Models are stored once in a content-addressed store. ComfyUI, A1111, and other tools see symlinks — no duplicate 24GB files.\n\n**GPU-aware.** Automatically picks the right model variant (fp16, fp8, quantized) for your VRAM. A 4090 gets full quality. An 8GB card still works.\n\n**Train LoRAs in one command.** Point it at a folder of images, pick a base model, and go. Powered by [ai-toolkit](https://github.com/ostris/ai-toolkit) under the hood, with auto-captioning, dataset prep, and sensible defaults included.\n\n---\n\n## Quick Start\n\n```bash\n# Install\ncurl -fsSL https://modl.run/install.sh | sh\n\n# Pull a model (auto-selects variant for your GPU)\nmodl pull z-image-turbo\n\n# Generate\nmodl generate \"a photo of a mountain lake at sunset\"\n```\n\nOr do everything at once:\n\n```bash\ncurl -fsSL https://modl.run/install.sh | sh -s -- --quick\n```\n\nThis installs modl, pulls a starter model, and launches the web UI.\n\n---\n\n## Web UI\n\n```bash\nmodl serve\n```\n\nGenerate, train, browse outputs, and manage models from the browser at `http://localhost:3939`. Same engine as the CLI.\n\n![modl web UI — generate tab](https://modl.run/ui-generate-lora.webp)\n\nInstall as a system service (starts on boot):\n\n```bash\nmodl serve --install-service\n```\n\n---\n\n## Train a LoRA\n\n```bash\n# Prepare dataset (auto-captions your images)\nmodl dataset create my-product --from ~/photos/product-shots/\nmodl dataset caption my-product\n\n# Train\nmodl train --dataset my-product --base flux-dev --name product-v1 --lora-type object\n\n# Generate with your LoRA\nmodl generate \"a photo of OHWX on marble countertop\" --lora product-v1\n```\n\n---\n\n## Supported Models\n\n16 models across 6 families. See the full comparison at **[modl.run/guides/model-comparison](https://modl.run/guides/model-comparison)**.\n\n| Family | Models | Best for |\n|--------|--------|----------|\n| **Flux 2** | Dev, Klein 4B, Klein 9B | Fast generation (4 steps), editing, best quality/speed |\n| **Flux 1** | Dev, Schnell, Fill Dev | Largest ecosystem, LoRAs, ControlNet, inpainting |\n| **Chroma** | Chroma | Apache 2.0, negative prompts, 8.9B Flux fork |\n| **Z-Image** | Base, Turbo | Strong quality/size, fast turbo, great ControlNet |\n| **Qwen Image** | Image, Image Edit | Text rendering (Chinese/English), instruction editing |\n| **Legacy SD** | SDXL, SD 1.5 | Low VRAM, massive LoRA library |\n\nPlus 70+ ControlNets, IP-Adapters, VAEs, text encoders, upscalers, and segmentation models. Browse all at **[modl.run/models](https://modl.run/models)**.\n\n```bash\nmodl pull flux2-klein-4b    # fast, 4-step generation + editing\nmodl pull flux-dev          # high quality, best for training\nmodl pull z-image-turbo     # strong quality, fast, great ControlNet\nmodl pull chroma            # open-source (Apache 2.0), negative prompts\n```\n\n---\n\n## Image Primitives\n\n### Generation \u0026 Editing\n\n```bash\nmodl generate \"prompt\" --base flux-dev          # text to image\nmodl generate \"prompt\" --init-image photo.png   # image to image\nmodl generate \"prompt\" --init-image img --mask mask.png  # inpainting\nmodl edit \"add sunglasses\" --image portrait.png  # instruction editing\n```\n\n### ControlNet \u0026 Style Reference\n\n```bash\nmodl preprocess canny photo.png                 # extract edges / depth / pose\nmodl generate \"prompt\" --controlnet edges.png   # structural control\nmodl generate \"prompt\" --style-ref painting.png # style transfer\n```\n\n### Vision-Language\n\n```bash\nmodl ground \"coffee cup\" cafe.png               # find objects → bounding boxes\nmodl describe photo.png                         # generate captions\nmodl vl-tag photo.png                           # auto-tag images\n```\n\n### Analysis \u0026 Post-Processing\n\n```bash\nmodl score photo.png                            # aesthetic quality (1-10)\nmodl detect photo.png                           # face detection\nmodl segment photo.png --bbox 120,340,280,500   # create masks (SAM)\nmodl face-restore photo.png                     # fix AI faces\nmodl upscale photo.png --scale 4                # 4x resolution\nmodl remove-bg photo.png                        # transparent PNG\nmodl compare ref.png target.png                 # CLIP similarity\n```\n\nEvery command supports `--json` for scripting and agent pipelines.\n\n---\n\n## Already Have Models?\n\n```bash\nmodl link --comfyui ~/ComfyUI\nmodl link --a1111 ~/stable-diffusion-webui\n```\n\nmodl scans your model folders, hashes files, and moves recognized models into the store — replacing them with symlinks. Your tools keep working, nothing breaks.\n\n---\n\n## Docker\n\n```bash\ndocker run --gpus all -p 3939:3939 -v modl-data:/workspace ghcr.io/modl-org/modl:latest\n```\n\nSet `MODEL=flux-schnell` to auto-pull a model on first boot. Models persist on the volume across restarts.\n\n---\n\n## Architecture\n\nSingle Rust binary for speed and distribution. Managed Python runtime for GPU compute. No external dependencies to install.\n\nFull CLI reference: **[modl.run/docs](https://modl.run/docs)**\n\n---\n\n## Author\n\nCreated by [Pedro Alonso](https://github.com/pedropaf).\n\n## License\n\n[AGPL-3.0](LICENSE)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmodl-org%2Fmodl","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmodl-org%2Fmodl","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmodl-org%2Fmodl/lists"}