SuperAPI AI Directory
Partner with us中文AI newsFavorites

AI images

3,482 results · Page 46 of 97

Stable diffusion XL

AI images

Stable Diffusion XL (SDXL) is a new version of the Stable Diffusion image generation system released by Stability AI in July 2023. According to the information, it uses a dual-model system with 6 billion parameters, supports 1024x1024 resolution, can generate highly realistic images and clear text, and simplifies prompt input with multiple built-in preset styles. Stable Diffusion itself is an open-source text-to-image tool based on diffusion models, developed by the CompVis team at the University of Munich and Runway ML, with computing support from Stability AI. It can perform image inpainting, outpainting, and text-guided image-to-image translation, and has open-sourced its code, pre-trained models, and license, allowing users to run it locally on a single GPU. Users can generate images through text prompts and customize them with Lora, embeddings, and negative prompts; it can be used online via web platforms or installed and run locally, and when run locally it can be fully offline. The information also introduces downloading, installation, training, and features such as ControlNet and face replacement, while noting limitations such as training data biased toward Western culture and poor generation of human limbs and faces.

Read more

StableCascade

AI images

Stable Cascade is the official codebase released by Stability AI, based on the Würstchen architecture, providing training and inference scripts as well as a variety of available models. Its core feature is that it operates in a highly compressed latent space: with a compression factor of 42, it can encode a 1024×1024 image into 24×24 while maintaining clear reconstruction, thereby bringing faster inference speed and lower training cost, making it suitable for image generation scenarios that value efficiency. The model consists of three parts: Stage A, Stage B, and Stage C. Stage A is a VAE, while Stage B and C are diffusion models. It is officially stated to support extensions such as fine-tuning, LoRA, ControlNet, IP-Adapter, and LCM, and performs well in prompt alignment and aesthetic quality evaluations. The code uses the MIT license, and the model weights follow the STABILITY AI NON-COMMERCIAL RESEARCH COMMUNITY LICENSE.

Read more

TalkPix: AI Talking Photo

AI images

TalkPix is an AI talking photo app developed by Ege Alp that turns static portrait photos into talking videos. Its core capability is to make faces move and achieve lip sync. Users only need to select a clear portrait photo and enter the words they want the person in the photo to say, and the app will use AI to convert the text into natural speech and generate a talking video in which the mouth movements match the speech. According to the official introduction, its lip-sync effect is relatively realistic, and the output is clear 720p video that can be downloaded to the photo album and shared to platforms such as TikTok and Instagram Reels. The app interface is simple, and operation requires only three steps: choose a photo, type, and tap generate. TalkPix is free to download on the App Store and includes in-app purchases. It is available for iPhone, iPad, Mac (requires Apple M1 chip or later), and Apple Vision devices. Its version 2.0 integrates the talkpix.ai website, supports sign-in with Apple, Google, or email, and the app and web share credits, videos, and voices.

Read more