Glitch-Embedding

EMBEDDING DOWNLOAD LINK – Glitch is a finetuned embedding inspired by 80s and 90s VHS tape aesthetics (trained on SD 2.1 768 ema pruned). With it you can style images overall and affect skin, clothing and the general appearance of people, animals and more.

Glitch generates both painted and photorealistic styles, where subjects and objects become more or less part of the VHS like glitch.

It works great with tv and film references,

and in an artsy, stylised sense to set a mood.

Install instructions and usage

1. Place either the glitch.pt or glitch.png file in the embeddings folder of your Automatic1111 installation.
2. Trigger the style in the prompt by writing glitch.

To get great results a very basic negative prompting is suggested:

ugly cartoon drawing, blurry, blurry, blurry, blurry

This negative prompt is used throughout images shown in this presentation, which is a shorter, edited version of Stability AI’s recommendation for SD 2.x. Turning on highres fix is also higly recommended to achieve the best results.

Example prompts and settings

TV/movie still:<br> glitch, close-up portrait of Millie Bobby Brown as Eleven, Stranger Things 1 9 8 2 movie still, Mitchell FC 65 Camera 35 mm, heavy grain<br> Negative prompt: ugly cartoon drawing, blurry, blurry, blurry, blurry<br> Steps: 30, Sampler: DPM++ SDE Karras, CFG scale: 7, Seed: 3729481696, Size: 1152x768, Model hash: 4bdfc29c, Denoising strength: 0.7

Dancer ripping up the glitch with her hand:<br> a close-up portrait of a wise Megleno-Romanians girl miner dancing in Spain, by glitch, 2d animation<br> Negative prompt: ugly cartoon drawing, blurry, blurry, blurry, blurry<br> Steps: 30, Sampler: DPM++ SDE Karras, CFG scale: 7.5, Seed: 888774459, Size: 1024x768, Model hash: 4bdfc29c, Model: sd_v2_v2-1_768-ema-pruned, Batch size: 5, Batch pos: 4, Denoising strength: 0.7

Credit

Thanks to masslevel who has contributed with a large number of images and knowledge on prompt settings. Thanks also to Klinter for providing the gif animation.

License

This model is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage. The CreativeML OpenRAIL License specifies:

1. You can't use the model to deliberately produce nor share illegal or harmful outputs or content
2. The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license
3. You may re-distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully) Please read the full license here