Three visual hypotheses
Each card uses the parent data, appearance prediction, coat/color possibilities, size range, and temperament notes to create a different realistic direction.
Saved Mixes
Compare Parents
Asset Credits
Current build
The main reveal is now a Hyperreal Mix Studio. It creates three breed-informed image directions for every mix: a balanced blend, a Parent 1-leaning version, and a Parent 2-leaning version. When local ComfyUI is reachable with a checkpoint, those three cards render real generated images. Without ComfyUI, the app still produces the full prompt set and visual reference cards.
Installed breed photo references
Parent and encyclopedia cards use local Wikimedia Commons reference photos when a license-tracked image is available. Missing breeds use the hyperreal prompt studio until a cleanly licensed photo is added to assets/photos/. Full metadata is stored in assets/breed-photo-credits.json.
| Breed | License | Source |
|---|---|---|
| Rottweiler | CC BY-SA 3.0 | Wikimedia Commons |
| Siberian Husky | CC BY-SA 3.0 | Wikimedia Commons |
| German Shepherd | CC BY-SA 2.5 | Wikimedia Commons |
| Golden Retriever | Public domain | Wikimedia Commons |
| Labrador Retriever | CC BY-SA 2.0 | Wikimedia Commons |
| French Bulldog | CC BY-SA 4.0 | Wikimedia Commons |
| Bulldog | CC BY-SA 4.0 | Wikimedia Commons |
| Beagle | CC BY-SA 3.0 | Wikimedia Commons |
| Poodle | CC BY 2.0 | Wikimedia Commons |
| Dachshund | CC BY-SA 4.0 | Wikimedia Commons |
| Boxer | CC BY-SA 3.0 | Wikimedia Commons |
| Cane Corso | CC BY-SA 2.5 | Wikimedia Commons |
Optional 3D fallback source
Source retained for fallback/experiments: Quaternius Ultimate Animated Animal Pack. It lists glTF support and a CC0/Public Domain license, making it the cleanest free 3D route if the app later needs a lightweight non-AI fallback again.
| Step | Action |
|---|---|
| 1 | Download the Quaternius pack from the source page. |
| 2 | Find a dog-like glTF or GLB in the glTF/GLB folder. |
| 3 | Rename the chosen entry file to dog-preview.gltf or update the viewer source. |
| 4 | Place it at assets/dog-preview.gltf. |
| 5 | Use it only as a fallback or supporting visual lane; the main reveal is the hyperreal three-variation image studio. |
CC-BY 3D fallback policy
Sketchfab can be used only when a model is downloadable and has clear CC-BY or CC0 terms. For CC-BY, record the exact model title, creator, source URL, license, download date, and modifications in assets/ATTRIBUTION.md. Avoid paid, unclear, non-downloadable, editorial-only, or restrictive assets.
"dog-preview.gltf" by [Creator] is licensed under CC-BY 4.0. Downloaded from: [URL]. Modifications: renamed file, compressed textures.
Optional local AI
ComfyUI is the intended free local photoreal workflow, but it is not bundled. Use sub-2GB SD 1.5 photoreal checkpoints where possible; any model or bundle larger than roughly 2 GB should be approved before installation. Generated images must remain labeled as visual approximations.
Prompt behavior: the app generates three prompts from the actual mix report, including size, coat, color, parent-leaning variation, temperament mood, and negative prompt controls for realistic anatomy.