v0.11.0 — Video generation (Wan 2.1), security hardening, gallery improvements

AI image & video generation
in your terminal.

Chewy is a beautiful TUI for generating images and videos with Stable Diffusion, FLUX, Wan, DALL-E, Imagen, and more. Run locally or connect to cloud providers — your choice.

Get Started View Source
$ brew install Holy-Coders/chewy/chewy
chewy
chewy | Local (sd.cpp) | dreamshaper-v8 SD 1.x [^y] provider [^n] models
Models
dreamshaper-v8
sd15-q8_0
flux1-schnell-F16
flux1-dev-Q2_K
realisticVision-v5
Prompt
a sunset over the ocean, vibrant colors, photorealistic
steps 20   cfg 7.0   512x512   seed -1   euler_a   karras   threads 4
Generated sunset image
output: outputs/20260313_012424.png | 25.4s | seed 42 ^e open | ^f fullscreen
Features

Everything you need.
Nothing you don't.

Run locally or connect to cloud providers. Full control over your image generation workflow.

🌐

6 Generation Providers

Local sd.cpp, OpenAI (DALL-E/GPT Image), Google Gemini (Imagen 3), HuggingFace, OpenAI-compatible APIs, and more.

🔒

100% Local Option

Run entirely on your hardware with stable-diffusion.cpp and Metal GPU acceleration. No cloud, no API keys, no tracking.

🎨

SD, FLUX, Wan, DALL-E & More

Stable Diffusion 1.x/2.x/3.5, SDXL, FLUX, Wan 2.1 video, DALL-E 3, GPT Image 1, Imagen 3, HiDream. Use .gguf, .safetensors, or .ckpt locally.

📥

Model & LoRA Browser

Curated starter models and LoRAs. Browse HuggingFace and CivitAI to download, search, and install without leaving the terminal.

🎬

Splash Screen & Custom Themes

Animated pixel-art splash screen on startup. 10 built-in color themes including Dracula, Catppuccin, Tokyo Night, Nord, and a light mode. Drop your own YAML theme files in ~/.config/chewy/themes/.

⚙

Samplers, Schedulers & Presets

14 samplers, 9 schedulers, configurable thread count, and built-in presets (fast, quality, portrait, flux-fast, flux-quality).

🖼

AI Prompts, Inpainting, LoRAs & More

AI-powered prompt enhancement using local LLMs (ollama) or cloud APIs. Click-to-paint mask editor for face-preserving inpainting. img2img, LoRAs with adjustable weights, and prompt history.

🎬

Video Generation (Experimental)

Generate videos with Wan 2.1 models via sd.cpp. In-terminal video player with play/pause, frame stepping, speed control, and optional MP4 export via ffmpeg.

🖥

Live Preview & Fullscreen

Real-time image preview as it generates. Fullscreen mode with Kitty graphics protocol. Auto model discovery from DiffusionBee and Draw Things.

Keyboard Shortcuts

Designed for speed.

Every action is a keystroke away. No mouse needed.

tab Cycle focus
enter Generate image
^y Switch provider
^n Model picker
^d Download models
^t Themes (custom supported)
^b Browse init image
^v Paste from clipboard
^a Gallery
^g Generation history
^l LoRA selection
^p Presets
^e Open in viewer
^f Fullscreen preview
^x Cancel generation
^r Randomize seed
^u Clear init image
^w Clear & start fresh
F1 Help overlay
^q Quit
Installation

Up and running in one command.

Chewy installs everything you need, including the sd inference engine with GPU acceleration.

Homebrew (recommended)

Installs chewy + stable-diffusion.cpp with Metal GPU support.

$ brew install Holy-Coders/chewy/chewy
$ chewy

From source

Requires Ruby 4.0+ and stable-diffusion.cpp on your PATH.

$ git clone https://github.com/Holy-Coders/chewy.git
$ cd chewy && bundle install
$ ruby chewy.rb

Configuration

Customize paths and defaults with environment variables.

# Set custom model directory
$ export CHEWY_MODELS_DIR=~/my-models

# Set custom output directory
$ export CHEWY_OUTPUT_DIR=~/Pictures/ai

# Point to a custom sd binary
$ export SD_BIN=/path/to/sd

Start generating locally.

Images and video, locally or via 6 cloud providers. One beautiful TUI for all your generation needs.

View on GitHub