OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Fizgig

Fizgig is an open-source Python workbench for training and fine-tuning image, video and music models with LoRA and LoKR techniques, and for profiling, repairing, exploring and extracting adapters.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Fizgig — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Fizgig when model training is only one part of the task and you also need a workbench for evaluating or adapting LoRAs. Its documented image workflow covers dataset setup through training, while additional tabs handle profiling, repair, exploration, epoch comparison and extraction.

Who it is for

  • Model creators training or fine-tuning supported generative models.
  • LoRA users who want to inspect, repair, compare or remix adapters.
  • Creators working with image, video or music model workflows.

Documented uses

  • Prepare image data, captions and preview prompts for a training run.
  • Train a LoRA, LoKR adapter or, for supported models, fine-tune the base model.
  • Profile adapter blocks and adjust them to repair a LoRA without retraining.
  • Compare training epochs using rendered previews.
  • Load compatible LoRAs from named training ecosystems and save .safetensors for ComfyUI.
Before you choose

Requirements & limits

The project describes Fizgig as free and open source under the Apache-2.0 license. Its README also describes deployment on rented hardware through RunPod.

  • Model support is specific: the README names Klein 9B, Krea 2, MiniMax H3, Qwen Image 2.1, Z-Image Turbo, SDXL, Anima and ACE-Step.
  • Compute needs depend on the model and task; the README gives examples of lower VRAM requirements for some workflows rather than one universal specification.
  • The documented image workflow expects a training-image folder and separates optional image preparation from captions and preview prompts.
  • The README names LoRA formats or ecosystems it can load, and identifies .safetensors output for ComfyUI.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up an image-training job

Prepare the documented inputs and settings for an image-model training run.

Before you start

  • A folder of training images.
  • A model and training preset to select in Fizgig.

Steps

  1. Set the training image folder in the Start tab.

  2. Optionally resize, face-crop or filter the images in Image Prep.

  3. Create trigger-word or AI captions in Captions.

  4. Set preview prompts in Samples.

  5. Choose a model and preset in Training, then click Start Training.

Expected outcome

A configured image-training run that produces training previews and a LoRA.

Guide limitations

  • The README does not give a single hardware requirement for every model and training task.
  • Image preparation is optional; the steps do not document a separate video or music training procedure.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
shootthesound/Fizgig · Official homepage title · retained in its source language
How we review sourcesSuggest a correction