Bytedance Inc.

Open-source agent harness that orchestrates sub-agents, memory, and sandboxes for tasks running minutes to hours.

Multimodal agent stack for controlling computers and browsers with natural language, via desktop app and CLI.

Handles debugging, refactoring, and feature work through an LLM agent that runs on multiple model providers.

ByteDance's native GUI agent model that reads screenshots and emits clicks and typing to operate desktop and web apps.

HTML5 video player whose built-in MP4/MSE parser loads only part of a file to cut bandwidth.

ByteDance workflow development framework with a built-in canvas and node-based primitives for designing AI workflows.

Open-source text-to-speech model from ByteDance offering multilingual voice synthesis and voice cloning.

ByteDance multimodal model with 3B active parameters for image and video understanding, generation, and editing.

Diffusion-based lip-sync model from ByteDance that syncs mouth movements in video to arbitrary audio.

Sandbox runtime from ByteDance for executing LLM-generated code safely across many languages during agent evaluation.

Batch-processing extension for ComfyUI that sweeps parameters across runs to speed up workflow debugging.

React library for visually building streaming step-by-step pipelines such as LLM chains and approval flows.

ByteDance model for photo recrafting that changes pose, scene, or style while keeping the same face identity.

Diffusion-transformer model from ByteDance research for high-fidelity face swapping in video.

Diffusion-transformer method from ByteDance for customizing single- and multi-subject image generation.

ByteDance model unifying style- and subject-driven image generation through disentangled and reward learning.

Model-based Android UI testing tool that learns app transitions to surface stability bugs.

Code for a NeurIPS 2025 method giving diffusion transformers consistent multi-subject identity and semantic control.

Portrait animation model using hierarchical motion attention to reenact expressive faces from a driving video.

Family of large-scale video-language models for video description and general video understanding.

Go mocking library from ByteDance that patches functions and methods at runtime for unit tests.

ByteDance's family of large multimodal models built for video understanding and editing tasks.

Android tool that pinpoints native memory leaks in C and C++ code from production apps.

Node for ComfyUI implementing HyperLoRA, parameter-efficient adaptive generation for portrait synthesis.