Analysis updated 2026-05-18
Upload a reference photo to get a text prompt that recreates a similar style image.
Generate Danbooru-style keyword tags from an image for AI art communities.
Reverse-engineer an existing image into a prompt you can tweak and remix.
Use lighter AI models to automatically caption folders of images on a modest GPU.
| adeliox/forge-neo-image2prompt | 0xustaz/streamgate | a-bissell/unleash-lite | |
|---|---|---|---|
| Stars | 1 | 1 | 1 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | hard | hard |
| Complexity | 3/5 | 4/5 | 4/5 |
| Audience | vibe coder | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires a Forge Neo installation and a GPU with at least 5 GB of video memory, first use triggers a multi-gigabyte model download from Hugging Face.
This repository is an extension for SD WebUI Forge Neo, which is a popular interface for generating images from text using AI. The extension adds a new tab called Image2Prompt. You upload an image, and the extension uses a vision-language model to describe that image in words. You can then send the generated text description directly into the image generation tools built into Forge Neo, creating a way to go from an image to a text prompt and back to a new image. The extension supports several AI models for understanding images. The recommended model is Qwen2-VL-2B-Instruct, which needs about 5 GB of video memory and offers a good balance between quality and resource usage. A larger 7B version exists for maximum quality but requires around 16 GB of video memory. There are also lighter Florence-2 models from Microsoft that focus on simple image captions. On first use, the chosen model downloads from Hugging Face, which can be several gigabytes. The extension caches these models so you only download once. You can choose how the output is formatted. Options include a natural Stable Diffusion prompt, Danbooru-style tags, or a detailed or short description. For the Qwen models, you can also write custom instructions to guide how the model describes your image. After generating a prompt, you get buttons to copy the text or send it directly to the txt2img or img2img tabs in Forge Neo. The extension also includes a button to free up video memory after each generation, which helps if your GPU has limited capacity. Installation involves copying the extension folder into the extensions directory of your Forge Neo setup and doing a full restart. On first launch, the extension checks for and installs required dependencies like transformers, accelerate, and einops. A graphics card is recommended for reasonable speed, though the extension does offer a forced CPU mode that works on any machine but runs slowly. The documentation includes a troubleshooting table for common issues like missing dependencies or running out of memory.
An add-on for the SD WebUI Forge Neo image generator that looks at any picture and writes a text description of it, which you can then feed back into Forge Neo to create a new image.
Mainly Python. The stack also includes Python, Transformers, PyTorch.
No license information is provided in the repository, so terms of use are unknown.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.