DreamLook

DreamLook is optimized for Stable Diffusion finetuning, letting you train models 2.5x faster and run thousands of jobs per day. The tool can be used to generate high-quality images using Stable Diffusion and its extensions, ControlNet and SDXL.

DreamLook is particularly useful for tasks like training models on pets, virtual photoshoots, and photorealistic images. It also can be used to extract LoRA files, which can shrink the size of downloaded checkpoints.

DreamLook's key features include:

• Unbeatable Dreambooth Performance: Train 1,200 steps in under three minutes without sacrificing quality.
• Lightning Fast Image Generation: Generate 1024x1024 Stable Diffusion images in 2.5 seconds.
• Reliable API: Run your applications without worrying about provisioning instances or finding GPU quotas.

DreamLook pricing is based on tokens, which can be used for both training models and generating images. Training a Stable Diffusion 1.5 model costs $0.75 per run, while SDXL training costs $3 per run. Image generation costs start at $0.01 per image for Stable Diffusion 1.5 and $0.04 per image for SDXL.

DreamLook's token packages include:

• 150 tokens: $15, includes up to 15 SD1.5 training runs, 3 SDXL training runs, and 1,200 generated images.
• 700 tokens: $60, includes up to 70 SD1.5 training runs, 17 SDXL training runs, and 5,600 generated images.
• 1,200 tokens: $100, includes up to 120 SD1.5 training runs, 30 SDXL training runs, and 9,600 generated images.
• 5,000 tokens: $400, includes up to 500 SD1.5 training runs, 125 SDXL training runs, and 40,000 generated images.
• 10,000 tokens: $750, includes up to 1,000 SD1.5 training runs, 250 SDXL training runs, and 80,000 generated images.

DreamLook offers a free subscription with 25 tokens when you sign up. It also offers more premium subscription tiers, including Enthusiast ($19/month), Pro ($199/month), and Enterprise (custom pricing), which include features like ControlNet support, higher maximum training steps and images, and dedicated queue priority.

Published on June 14, 2024