LAION AI ORG

CLIP-like model evaluation

Contrastive Language-Audio Pretraining

No description.

Aim for the moon. If you miss, you may hit a star.

Projects at LAION

OpenAssistant is a chat-based assistant that understands tasks, can interact with third-party systems, and retrieve information dynamically to do so.

Audio Dataset for training CLAP and other models

A linear estimator on top of clip to predict the aesthetic quality of pictures

Pretrained Dalle2 from laion

No description.

No description.

A summary of all lucidrains repositores and links to training / research approaches by LAION or other communities.

Collect large 3d dataset and build models

Description and pointers of laion datasets

A phenaki reproduction using pytorch.

Open Instruction Generalist is an assistant trained on massive synthetic instructions to perform many millions of tasks

Reproducible scaling laws for contrastive language-image learning (https://arxiv.org/abs/2212.07143)

Home of `erlich` and `ongo`. Finetune latent-diffusion/glid-3-xl text2image on your own data.

No description.

Alice in Wonderland code base for experiments and raw experiments data

No description.

Implementation of a discord channel scraper to generate datasets.

Let's make a video clip

O-GIA is an umbrella for research, infrastructure and projects ecosystem that should provide open source, reproducible datasets, models, applications & safety tools for Open Generalist Interactive Agents (O-GIA). O-GIA systems will act in collaboration with human or autonomously, supporting various kind of validated decision making and assistance.

A repository containing datasets and tools to train a watermark classifier.

An open toolbox for NSFW & toxicity detection

No description.

No description.

Big-Interleaved-Dataset

No description.

Tools for content datamining and NLP at scale

BUD-E (Buddy) is an open-source voice assistant framework that facilitates seamless interaction with AI models and APIs, enabling the creation and integration of diverse skills for educational and research applications.

No description.

Adversarial Training and SFT for Bot Safety Models

No description.

Inverts CLIP text embeds to image embeds and visualizes with deep-image-prior.

Building the laion5B paper

This repository contains prompts & best practices to annotate audio clips with a very high degree of details using Audio-Language-Models

Aggregating embeddings over time

This repository will be a summary and outlook on all our open, medical, AI advancements.

(wip) Use LAION-AI's CLIP "conditoned prior" to generate CLIP image embeds from CLIP text embeds.

Anh - LAION's multilingual assistant datasets and models

Scaled diffusion transformer for text-to-speech synthesis (DiT + T5Gemma2 conditioning, TorchTitan & Megatron backends, tested up to 1024 GPUs)

BUD-E (Buddy) is an open-source voice assistant framework that facilitates seamless interaction with AI models and APIs, enabling the creation and integration of diverse skills for educational and research applications.

Un-*** 50 billions multimodality dataset

No description.

A frontend that is compatible to the school-bud-e-backend.

Here we provide and collect many functions to generate math problem and step by step solutions for LLM training

No description.

A general human-ai interaction platform.

Open-weights voice acting pipeline combining zero-shot voice cloning with natural-language direction. Provide a reference voice (or generate one) and describe how the line should be performed. Produces speech that keeps the voice identity while following emotional and stylistic prompts—no training required.

No description.

Frontend (and soon also midleware and backend) for a new, opensource image generation platform.

This project provides a data set with bounding boxes, body poses, 3D face meshes & captions of people from our LAION-2.2B. Additionally it provides clusters based on the poses and face meshes and pose-related captions based on these cluster assignments.

No description.

This is the LAION repository for creating open super-resolution models with the help of LAION-5B subsets.

Describe the format of image/text datasets

This repository contains training code and checkpoitns for finetuning glide.

A collection of generative and training notebooks getting mirrored to google colab.

Easily compute model embeddings and save the embeddings.

Official repo for Project Alexandria

No description.

No description.

Knowledge Acquisition and Interlinking via Semantic Embeddings and Reasoning

MegaTron open-sci fork

laion github website

This repository is a summary of all systems and scientific papers that use LAION datasets.

Collection of three complementary voice taxonomies: VoiceNet (59 speech dimensions), EmoNet (40 emotion categories), VocalBurst (82 non-speech sounds)

Retrieval-augmented voice cloning and emotion conditioning data generation pipeline. Combines Echo TTS, ChatterboxVC, and Empathic Insight Voice+ to generate large-scale datasets of emotionally conditioned speech with disentangled speaker identity and emotional prosody.

OpenCLIP fork with MaMMUT support

A collection of safety classifiers and models to process image and texts.

This repository will give a quick overview of all projects and repositories from LAION.

No description.

LaionBox: Fine-tuned DramaBox TTS with Multi-Auxiliary Differentiable Losses

Official repository for the NeurIPS 2025 paper “EmoNet-Face: An Expert-Annotated Benchmark for Synthetic Emotion Recognition.” Includes a 40-category emotion taxonomy, balanced synthetic datasets, expert annotations, and baseline models for fair and reproducible evaluation.

Experiments with Summarization, Long Context and Retrieval

A client library for Crawling@Home's effort to filter CommonCrawl with CLIP, building a large scale image-text dataset.

PyTorch code for BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation

No description.

A repository with data for annotation.

A basic setup for decentralized-learning that can be used for training future DALLE/CLIP/CLAP models.

This repository shall help finding a good distribution for huge datasets like LAION-5B for more efficient training.

The new repository for the genral inference pipeline.

General / Global Inference Framework

This is the development repository of the LAION-AI website.

No description.

Detect & caption vocal bursts (laughs, sighs, gasps, screams...) — laion/vocalburst-locator + laion/vocalburst-captioning-whisper ensemble. Inference code, weights, benchmark & demo.

datasets that should be downloaded & converted to our standard training formart.

High-level Python library for zero-shot voice conversion using Resemble AI's Chatterbox S3Gen model

Building an agentic voice assistant for mobile & desktop devices with episodic, semantic & procedural memories

A template for procedural template generation using JSON outputs form LLMs.

Alice in Wonderland project and initiative webpage

Creating subsets from laion5b via embeddings search

No description.

File server for curiosit-e content.

No description.

A prototype showing how to stream using Django x htmx.

A model to run hand inference on a cluster.

A repo containing images for artifact annotation.

A collection of public domain images donated for ML training.

No description.

All media / publicity on LAION and related stuff!

Recommended intro resources

Self-contained DramaBox voice acting pipeline: VoiceNet taxonomy, multi-GPU prompt generation, TTS synthesis, and audio refinement

School Bud-E is an intelligent and empathetic learning assistant designed to revolutionize the educational experience.

Admin Bud-E is a lightweight, privacy-first control center for AI chat, speech-to-text, and text-to-speech. Manage providers, routing, and costs with a simple Admin Console. Give users per-period credits, prices per model, and a shared Common Pool. EU-friendly via OpenAI-Format endpoints or our optional Google Cloud Vertex proxy.

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

No description.

Mobile app development of all bud-e derivatives.

This repository is a collection of found biases in the LAION-5B dataset.

Audio -> DramaBox performance prompts: procedural (Empathic-Insight-Voice-Plus) and ensemble (whisper experts + vocal bursts -> Gemma-4-4B) paths.

User interaction data is mostly opaque and saved in big tech companies (Claude, OpenAI, Google). This project tries to engage the public to use v1-compatible endpoints and frontend chats to donate its user interaction data for the development of future open frontier models.

No description.

Benchmark analysis

JAX/TPU training code for EchoTTS with DACVAE latent codec

No description.

No description.

Multi-node scaling benchmarks for CLAP contrastive audio-language models on HPC clusters

No description.

No description.

No description.

No description.

No description.

Bloom Server for Crawling@Home

No description.

This repository is wip website for lasting, open AI art.

This is a testing site for the redesign of the LAION.AI website.

Tracking compute allocation and compute donation

This repository provides access to human labeled datasets for training classifiers.