🔥 Black Forest Labs’ FLUX1.1 [pro] breaks new ground
AI image creation just got a major upgrade – here’s how
Total reading time around 4 minutes.
Welcome to Visually AI!
🔮AI News this Week
FLUX1.1 [pro] & BFL API
Black Forest Labs announced the updated FLUX1.1 [pro] model and the BFL API.
FLUX1.1 [pro] (codename “blueberry”) is their most advanced and efficient image model, with improved prompt adherence, image quality, and diversity in results. It’s also twice as fast, compared to previous models.
They partnered with:
There are also several apps and platforms that quickly added the new FLUX1.1 [pro].
The results are impressive - I generated these on Replicate’s Flux Gallery:
I think FLUX models have been incredible for developers and consumers as high-quality, open source image generators. We have more choices of apps and platforms to use for great images that rival Midjourney.
OpenAI Dev Day
OpenAI held Dev Day 2024, with several announcements, including:
Realtime API: Developers can now build fast, multi-modal conversational experiences with text and audio input and output, with adjustable natural voices with emotion.
Vision Fine-Tuning: Improve performance of GPT-4o for image tasks with as few as 100 training images and text.
Prompt Caching: Get automatic discounts on repetitive system prompts, common instructions, and other prompts to reduce cost by 50% and improve latency.
Distillation: Use outputs from a large model to fine-tune smaller models for similar performance on certain tasks and improve efficiency.
📸AI Snapshots
Pika is back with a new and improved video model and “Pikaffects” to squish, explode, cake-ify, and more. Camera motion control, Styles, and Effects can be included in the text prompt. The web app was overwhelmed quickly after Pika 1.5 launched with generations taking several hours - they’re working on it, though.
Here are a few image to video examples from Pika’s community gallery:
Hedra launched it’s newest foundational model, Character-2 with improved rendering and choice of aspect ratio (square, portrait, landscape). You can upload an image and voice file, or generate an image with Stable Difffusion, Flux Schnell, or Flux + Realism.
One of my results using the Flux Schnell model on Hedra:
Fal AI added video models to its platform (Gen-3, Luma Dream Machine & Kling), allowing you to run workflows with different models. You don’t need a separate API key for each video model - pay for the generations through Fal:
Gen-3: $1 ~ 4 generations
Luma: $1 ~ 2 generations
Kling: $1 - 33 generations (not a typo!)
Hailuo x MiniMax launched their image to video feature for all users. Prompt adherence and video quality is impressive. It does a good job with or without prompts, too.
Some of my results:
🛠️ This Week’s AI Tools
Magic Patterns: Prototype your product ideas with AI, import designs from the internet and prompt for your idea in the interactive generative UI canvas. (link)
LlamaCoder: Turn ideas into apps with a simple text prompt in minutes. Built with Llama 3.1 and Together AI. (link)
Cabina: AI Co-pilot platform that also lets you compare LLMs and AI image models. (link)
LipDub: AI lip sync tool to animate characters with voices in 40+ languages. (link)
Replicate Flux Gallery: Generate and compare image results on a canvas with all 4 FLUX models, including the latest FLUX1.1 [pro]. (link)
Expression Editor: Edit facial expressions in real time on fffiloni’s Hugging Face demo for fofrAI’s model based on LivePortrait and advanced ComfyUI custom nodes. (link)
Example of Expression Editor:
🖼️ Image Prompts
Prompt: Masterpiece [athletic brand] commercial, a [type of model] model is [athletic activity] with [athletic brand] sports wear and [brand apparel], in high speed
Example: Masterpiece Nike commercial, a female model is running with nike sports wear and nike sneakers, in high speed
Prompt: Futuristic garden with metallic flowers blooming under an artificial sun, their petals intricate and sharp; vibrant colors clash with the cold, reflective surfaces of robotic insects hovering above, casting elongated, surreal shadows
🎬Video Prompt
Prompt: Raindrops falling on a leaf in a lush green forest. Extreme close-up, vivid colors.
Thanks for reading and have a creative week!