Roadmap
Where Gentastic is headed
No fixed dates - this is a small, actively-developed project. Here's what's shipped, what's next, and what's on the horizon.
Timeline
Core text-to-image & image-to-image Shipped
FLUX.1 (schnell & dev) and FLUX.2 klein, with automatic Hugging Face downloads, license gating, and CUDA/Vulkan runtime auto-detection.
SDXL model family Shipped
Photoreal and anime-focused SDXL checkpoints alongside FLUX, each with automatic quality-tag prefixes and per-model resolution presets.
Keep-face identity generation Shipped
PhotoMaker-based identity lock for SDXL, plus face-preserving instruction editing on FLUX Kontext & klein - generate or edit a specific person from one reference photo, no training required.
Instruction-based image editing Shipped
FLUX.1 Kontext and FLUX.2 klein editing - describe a change and it's applied while the rest of the image is preserved.
Gallery, presets & on-demand CUDA Shipped
Metadata-rich gallery with a disk thumbnail cache, prompt/parameter presets, and a no-toolkit-required on-demand CUDA runtime download.
Responsible-use guardrails Shipped
Opt-in age gate for adult models, plus a per-generation acknowledgement of the rules - before every adult or identity-locked generation.
Native AMD ROCm 7 In progress
Vulkan already covers AMD GPUs today. The upstream StableDiffusion.NET ROCm backend is being integrated and validated on real AMD hardware before Gentastic bundles it.
Packaged installer & auto-update Planned
A signed installer and in-app update checks, on top of today's portable ZIP.
Inpainting, upscaling & ControlNet Planned
Targeted region editing, quality upscaling, and structural guidance for more precise control over composition.
Video generation Planned
Local, GPU-accelerated video generation is on the roadmap as the next major surface for Gentastic - same philosophy: pick a model, describe what you want, no workflow graphs.
Have an idea, or found a bug?
Gentastic is open source - issues and pull requests are welcome.