Crackio's prompt enhancer
LoRA for expanding text prompts
2 min readupdated
This compact LoRA (a small fine-tuning add-on for the base model) is built for the LiquidAI/LFM3-VL-2B model and acts as a detailed description generator for images and video. It preserves all details of your request, fixes typos, translates other languages into English, and adds environmental elements, camera angles, lighting, and mood without hallucinating extra details. The model was trained on 6,041 example pairs and designed as a built-in tool for the Crackio studio, but it can also be used in other projects on your own GPU.
Key features
- Precise preservation of original details without adding new objects, brands, or people.
- Automatic addition of framing, lighting, tone, and mood settings.
- Support for translating other languages into high-quality English prompts.
- Addition of natural camera motion or behavior for video generation.
What you can do
- Use it in the open Crackio studio to quickly enhance any text prompts before generation.
- Apply it as a local translation and idea refinement tool on your own GPU.
- Integrate it alongside the base LFM3-VL-2B model in your own graphics pipelines.
- Create detailed prompts for video generators with specified camera behaviors.
Example
Run the model with the system prompt from the system.txt file and send your request in the Request: format to get an expanded prompt in English.
Read more
DetailsBenchmarks: 97.3 out of 100 points based on reviewer ratings on test sets
Requirements: works on a GPU like the RTX 3090, LoRA file size is about 131 MB
License: LFM Open License v1.0 (commercial use allowed with revenue up to $10M per year)
Limitations: outputs text in English only, occasionally translates words with errors










Comments