Practical overview
When it fits
Use it when you want script-based local image translation with a documented paired or unpaired workflow, or when you plan to adapt the model using task-specific data.
Who it is for
- Developers who want to run local image-to-image inference from Python scripts.
- Researchers and builders adapting pretrained diffusion models to new image domains with custom data.
- Users exploring paired sketch or edge translation and unpaired visual changes such as day-to-night.
Documented uses
- Turn a sketch and prompt into a rendered image.
- Generate an image from Canny edges and a prompt.
- Translate an image between domains such as day and night or clear and rainy.
- Adapt a pretrained single-step diffusion model to a chosen task or domain using custom data.
Requirements & limits
The repository is public, written in Python and licensed under MIT. Its README documents local inference and a locally hosted Gradio demo.
- Paired workflows take an image and prompt; the documented unpaired day-to-night command takes an image and a selected translation direction.
- The repository supplies both a Conda environment file and an alternative virtual-environment installation path.
- The project describes single-step inference and custom-data adaptation as distinct uses.
- A locally hosted Gradio demo is described; the repository does not document an existing network-accessible API.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Run a paired sketch-to-image example
Set up the Python environment and run the repository’s documented sketch-to-image inference command.
Before you start
- Python environment using the repository’s Conda environment file or virtual-environment requirements
- An input sketch image
- A prompt and output directory
Steps
Create the Conda environment with
conda env create -f environment.yaml, then activate it withconda activate img2img-turbo; alternatively, create and activate a virtual environment and installrequirements.txt.Run the command below.
python src/inference_paired.py --model_name "sketch_to_image_stochastic" --input_image "assets/examples/sketch_input.png" --gamma 0.4 --prompt "ethereal fantasy concept art of an asteroid. magnificent, celestial, ethereal, painterly, epic, majestic, magical, fantasy art, cover art, dreamy" --output_dir "outputs"
Expected outcome
The script is documented to save a generated image in the specified output directory.
Guide limitations
- The documented command uses the repository’s example sketch, prompt and gamma value; the README does not specify additional hardware requirements in this setup example.
Run a day-to-night image translation
Set up the Python environment and invoke the documented unpaired day-to-night transformation.
Before you start
- Python environment using the repository’s Conda environment file or virtual-environment requirements
- An input day image
- An output directory
Steps
Create and activate the Conda environment with
conda env create -f environment.yamlandconda activate img2img-turbo, or set up the virtual environment and installrequirements.txt.Run the command below.
python src/inference_unpaired.py --model_name "day_to_night" --input_image "assets/examples/day2night_input.png" --output_dir "outputs"
Expected outcome
The script is documented to save a night-style translation of the input image in the specified output directory.
Guide limitations
- The documented command demonstrates the repository’s example input; the README does not detail how to train or configure a different unpaired translation task in this command.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- GaParmar/img2img-turbo · Official homepage title · retained in its source language