Proteus

Proteus serves as a sophisticated enhancement over OpenDalleV1.1, leveraging its core functionalities to deliver superior outcomes. Key areas of advancement include heightened responsiveness to prompts and augmented creative capacities. To achieve this, it was fine-tuned using approximately 220,000 GPTV captioned images from copyright-free stock images (with some anime included), which were then normalized. Additionally, DPO (Direct Preference Optimization) was employed through a collection of 10,000 carefully selected high-quality, AI-generated image pairs.

In pursuit of optimal performance, numerous LORA (Low-Rank Adaptation) models are trained independently before being selectively incorporated into the principal model via dynamic application methods. These techniques involve targeting particular segments within the model while avoiding interference with other areas during the learning phase. Consequently, Proteus exhibits marked improvements in portraying intricate facial characteristics and lifelike skin textures, all while sustaining commendable proficiency across various aesthetic domains, notably surrealism, anime, and cartoon-style visualizations.

ProteusV0.5 Update Notes

This update marks a significant leap forward, featuring a substantially trained custom CLIP model - a rarity in the field of AI image generation. This advancement, alongside enhanced stylistic capabilities, pushes Proteus closer to Midjourney-level performance while maintaining its unique strengths.

Key Improvements in V0.5:

1. Advanced Custom CLIP Integration:

   • Incorporates a meticulously trained custom CLIP model

   • Steadily developed over an extended period

   • Further fine-tuned for specific qualities in Proteus and Prometheus

   • Estimated to contribute 90% of the model's performance improvements

   • Requires a clip skip setting of 2 for optimal performance

   • Estimated to be responsible for 90% of the improvements in this version

2. Further Refinement of Stylistic Capabilities:

   • Enhanced ability to generate diverse artistic styles

   • Improved coherence in complex scenes and compositions

3. Expanded Training Dataset:

   • Now totaling over 400,000 images

   • Significantly broadened knowledge base and generation capabilities

4. Balanced Creativity and Accuracy:

   • Addressed previous issues of being "too stylistic/creative"

   • Improved alignment between user prompts and generated outputs

All improvements have been achieved using methods that do not infringe on any copyrighted material.

Settings for Proteus

Use these settings for the best results with ProteusV0.5:

Clip Skip: 2

CFG Scale: 7

Steps: 25-50

Sampler: DPM++ 2M SDE

Scheduler: Karras

Resolution: 1024x1024

ProteusV0.5 Update Notes

This update enhances stylistic capabilities, similar to Midjourney's approach, while also improving prompt comprehension. Methods used do not infringe on any copyrighted material.

Key improvements in V0.5:

• Further refinement of stylistic capabilities

• Enhanced prompt comprehension through a custom-trained CLIP model

• Improved balance between creativity and accuracy

• Expanded training dataset, now totaling over 400,000 images

The model has been fine-tuned and trained on a total of 400k+ images at this point, significantly expanding its knowledge base and capabilities.

Previous issues noted in V0.4 regarding "being too stylistic/creative" have been addressed in this release. V0.5 aims to strike a better balance between creative output and accurate representation of prompts.