Open source

Faster inference and on-device deployment are shaping how open-source image and video tools reach production this week.

• 1

  Ideogram_4_turbotime_loraHF Models

  ostris· 97 likes

  A LoRA adapter for Ideogram 4 that enables inference in as few as 2 steps without classifier-free guidance or an unconditional model pass. Builders looking to reduce generation latency in Ideogram 4 pipelines have a ready-made diffusers-compatible starting point, with updated versions planned as the underlying continuous turbo training method matures.

  diffuserstext-to-imagelorabase_model:ideogram-ai/ideogram-4-fp8base_model:adapter:ideogram-ai/ideogram-4-fp8
• 2

  LTX 2.3HF Models

  Lightricks· 1,424 likes

  LTX-2.3 is a 22-billion-parameter DiT-based model from Lightricks that generates synchronized video and audio together in a single pass, supporting image-to-video, text-to-video, and several other conditioning modes. Builders working on multimodal generation pipelines may find the distilled 8-step checkpoint particularly practical for iteration, given its reduced inference cost relative to the full model.

  diffusersimage-to-videotext-to-videovideo-to-videoimage-text-to-videoaudio-to-video
• 3

  Z Image TurboHF Models

  Tongyi-MAI· 4,848 likes

  Z-Image-Turbo is a text-to-image model built on a single-stream diffusion transformer architecture, designed for efficient image generation from natural language prompts. With over 930,000 downloads and native diffusers integration via a custom ZImagePipeline, it is a practical option for builders looking to deploy fast image generation on Azure or integrate it directly into existing diffusers-based workflows.

  diffuserssafetensorstext-to-imageendiffusers:ZImagePipelinedeploy:azure
• 4

  runanywhere-sdksGitHub

  RunanywhereAI· 10,316 stars

  A production-focused C++ SDK for running LLMs, speech-to-text, and text-to-speech locally on iOS and Android, with no network dependency. Builders targeting mobile or edge deployments can use it to ship on-device inference without relying on cloud APIs.

  androidapple-intelligencecppdiffusion-modelsedgeflutter
• 5

  krita-ai-diffusionGitHub

  Acly· 10,232 stars

  A Krita plugin that integrates Stable Diffusion for in-canvas image generation, inpainting, and outpainting with minimal configuration. Builders working on artist-facing AI tooling can study how it handles seamless ComfyUI/Automatic1111 backend integration within a professional painting application.

  generative-aikrita-pluginstable-diffusion