For best outputs: Sampler: DPM++ 2M Karras, Euler a. Steps: 12, cfg 1 ~ 3.5, CLIP skip 2. VAE: none required. Pos prompt: this model emulates SD3. It's capable from perfect photorealism to anime, 3D, cinematic, knows all possible art styles, it's great at celebrities knowledge, pop art, cinema culture, and it's really good at creating text. In order to text, put your text between simple ', for example, sign saying 'hello world'. It helps the model if the text goes written onto a surface, from a sign to a bottle or a board. Floating text may happen, but it's harder, unless you use a movie logo or similar. The model listens really well, don't overuse big weights. Neg prompt: you can leave it blank or simply what you don't want to see. Note about the speed: this model generates 3 outputs in average 12 seconds. Checkpoint Trained.