H3 can also be used as an image edit model and its surprisingly good. V1 - A simple image edit<ul><li>For V1 there are no additional requirements outside of regular minimax models</li></ul>V2 - Image editing with references<ul><li>V2 uses an <a target="_blank" rel="ugc" href="https://huggingface.co/Mamad8/MiniMax-H3-Image-VAE/tree/main">experimental VAE</a> which is tailored towards Image gen</li><li>Up to nine reference images can be used</li></ul> Prompting Tell the model what you want to change / add / remove, then tell it what to keep.<pre><code>Change the wave explosion in the background into an erupting volcano. Preserve the composition, lighting, style, facial expressions and backgrounds.</code></pre> Resolution Note that H3 only accepts specific image sizes natively, if your image is of a different size/resolution it is automatically rescaled but the result might have a difference in quality.<ul><li>608 x 352</li><li>736 x 416</li><li>864 x 480</li><li>960 x 544</li><li>1056 x 608</li><li>1152 x 640</li><li>1216 x 672</li><li>1280 x 736</li><li>1344 x 768</li><li>1376 x 768</li><li>1504 x 832</li><li>1664 x 928</li><li>824 x 1024</li><li>1920 x 1088</li></ul>Performance At 1280p it takes me around 24 seconds, I'm using an RTX 5080 16gigs and 64gigs RAM.

Files with SHA256: 19113c8e6aa64f1da790556873d7acb72ff16b663bbfc6fba0f11aee22508589