Week of 3 August 2026
This week's picks focus on reducing setup friction for local image and video generation workflows.
- 1KromaHF Models
lodestones· 148 likes
Kroma is a LoRA fine-tune for Krea 2, distributed as a single ComfyUI-compatible safetensors file that combines 264 low-rank adapters (rank 256) with full weight deltas for RMSNorm and modulation tensors to faithfully reproduce the fine-tuned behavior. Builders who already run Krea 2 in ComfyUI can drop this into their existing loras folder and load it via the standard LoraLoader node without any custom code or format handling.
lorakrea2kreatext-to-imagecomfyui - 2PrunaVAEDHF Models
PrunaAI· 59 likes
PrunaVAED is a pruned VAE decoder component built on the LTX 2.3 architecture, designed for text-to-video pipelines using the diffusers library. Builders looking to reduce inference overhead in LTX-based video generation workflows may find it a practical drop-in replacement for the standard decoder.
diffuserssafetensorsvideovaedecoderltx - 3awesome-gpt-image-2GitHub
YouMind-OpenLab· 9,089 stars
A curated, daily-updated library of 2,000+ prompts for OpenAI's GPT Image 2 model, organized with preview images and support for 16 languages. Builders exploring the model's text rendering or cross-image consistency capabilities can use it as a reference for what kinds of prompts reliably produce production-quality results.
ai-image-generationai-promptsawesomeawesome-listcommercial-illustrationduct-tape - 4ABot World 0 5B LFHF Models
acvlab· 67 likes
ABot-World-0-5B-LF is a 0.5B-parameter image-to-video world model built on the diffusers library that generates interactive video rollouts from a single input image. Builders working on interactive or game-like environments may find it worth evaluating given its stated ability to run on a single desktop GPU, which lowers the infrastructure bar for local experimentation.
diffuserssafetensorsti2vworld_modelabot-worldimage-to-video - 5StabilityMatrixGitHub
LykosAI· 8,587 stars
StabilityMatrix is a cross-platform package manager that handles installation and management of popular Stable Diffusion interfaces, including Automatic1111's sd-webui and ComfyUI. Builders who want to run multiple inference backends side by side without manual Python environment conflicts will find the unified launcher useful.
aiautomatic1111avaloniacomfyuidotnetpython