dreamlook.ai

Dreamlook.ai is an AI-powered finetuning platform for Stable Diffusion designed to accelerate training and scale experiments. It promises end-to-end training capabilities that go beyond LoRA adaptations, enabling full model finetuning and high-resolution image generation for SD1.5 and SDXL. The service emphasizes speed (training thousands of runs per day), scalability, and the ability to extract LoRA files to reduce download sizes, while offering an API to power production-grade applications without managing GPU infrastructure.

---

What it does

• Finetune Stable Diffusion 1.5 and Stable Diffusion XL (SDXL) with full model training, not just LoRA adapters.
• Train up to 1,000s of runs per day with optimized performance claims (e.g., 1,500 SDXL steps in 10 minutes, 4-second full-res SDXL image generation).
• Extract LoRA files from trained models to minimize file sizes for deployment and sharing.
• Create high-quality, high-resolution SDXL images (even up to 1024x1024) with precise control using tools like ControlNet, Offset Noise, and domain-specific base models.
• Offer a scalable Finetuning API so developers can integrate training and inference into their applications without managing GPU quotas.
• Provide a tiered pricing model for finetuning and generation, with examples like per-run and per-image costs across SD1.5 and SDXL.
• Support for training on various subjects (e.g., pets) and multiple base model options (30+ base models).
• Documentation, API access, and community resources (FAQs, testimonials, Discord, social links).

---

How to Use Dreamlook.ai

1. Sign up for free (no credit card required) and choose a pricing package.
2. Upload your training data (images) and configure target styles, objects, or poses.
3. Start full-model finetuning or generate using the optimized SD1.5/SDXL pipelines.
4. Compare base models and adjust training parameters to refine results.
5. If needed, extract LoRA files from your trained model for lightweight deployment.
6. Use the API to integrate training and generation into your apps or workflows.

Safety and Privacy

• The platform processes uploads for training and generation; check policies for data usage and privacy.

Pricing (Selected Examples)

• Finetuning Stable Diffusion 1.5 starts at $0.75 per run; SDXL starts at $2.25 per run.
• Image generation costs: SD1.5 from $0.01 per image; SDXL from $0.02 per image.
• Packages include various quotas for training runs, training jobs, and generated images (e.g., $15, $60, $100, $400, $750 with increasing allowances).

FAQ Highlights

• API availability for training and generation: Yes.
• Ability to download trained model checkpoints: Yes.
• SDXL availability: Yes.
• LoRA training and extraction: Supported.
• Training on objects or styles: Supported.
• Compatibility with AUTOMATIC1111 and RunDiffusion/ThinkDiffusion: Confirmed.
• Offset Noise feature: Available.
• Data handling: Details provided in privacy policy.

Testimonials

• Positive feedback focusing on speed, quality, and ease of use from users.

Core Features

• Full model finetuning (not just LoRA) for SD1.5 and SDXL
• Training at scale: 1,000s of runs per day with fast iteration
• High-speed SDXL image generation (rapid renders)
• LoRA extraction to reduce download sizes
• ControlNet and advanced image control features (e.g., Offset Noise)
• Access to 30+ base models and diverse subject support (pets, etc.)
• API access for integration into apps and pipelines
• Easy sign-up with free tier and transparent pricing
• Documentation, FAQs, and community support