• Ainsider
  • Posts
  • Ainsider #22: Insane AI developents

Ainsider #22: Insane AI developents

Adobe Premiere with Gen AI | Llama3 | Stable Diffusion 3 | Latest AI Tools

Ainsider #22

Adobe is changing Video production forever

Huge AI developments in last week - Meta, Microsoft, Stability AI, Grok, Krea

Latest AI Tools i found

Tremendous update of Adobe Premiere Pro - with GenAI

Adobe showed demo of their new Adobe Firefly video model - inside of Premiere Pro App 📹

What's even more exciting here - you'll be able to include third party Video platforms integration like Runwayml, Pika Labs, and OpenAI’s Sora 🤯

Stability AI released Stable Diffusion 3 model

Here's everything you need to know & key features of the Stable Diffusion 3 model 🧵👇

✅Improved Text Generation

Stable Diffusion 3 has significantly better text rendering capabilities compared to previous versions, allowing it to generate long sentences with accurate font styles and spacing.

It outperforms state-of-the-art text-to-image models such as DALL·E 3, Midjourney v6, and Ideogram v1 in terms of typography and prompt adherence, confirmed by human preference evaluations.

✅Enhanced Prompt Following

The model uses highly accurate image captions in training to improve its ability to follow prompts, making it as good as or better than DALL-E 3 in this regard.

✅Transformer-based Architecture

Stable Diffusion 3 utilizes a new Multimodal Diffusion Transformer (MMDiT) architecture, which is more efficient and scalable than previous models while also producing higher-quality images.

✅Rectified Flow Sampling

The model employs a rectified flow technique that simplifies the generative process, allowing for faster and more cost-effective image creation.

✅Availability and Pricing

Stable Diffusion 3 is currently available through the Stability AI Developer Platform API, with plans to make the model weights available for self-hosting with a Stability AI Membership in the near future.

The model is not free, but the membership fees are reported to be quite modest for small businesses.

Stable Diffusion 3 represents a significant advancement in text-to-image generation, especially with their API - which is a great starter for building own AI image models.

Meta Llama 3 model released

Everything you need to know 👇

✅8 billion and 70 billion parameter pre-trained and fine-tuned models available openly

✅Trained on over 15 trillion tokens, 7 times larger than the dataset used for Llama 2

✅Improved tokenizer with a vocabulary of 128,000 tokens for better performance

✅State-of-the-art performance across industry benchmarks

✅New capabilities such as enhanced reasoning and coding

✅3 times more efficient training compared to Llama 2

✅New trust and safety tools including Llama Guard 2, Code Shield, and CyberSec Eval 2

✅Integration into Meta AI

✅Availability in more countries across the Meta AI Apps

New Multimodal very powerful LLM

It's significant advancement for artificial intelligence 👀

Reka has introduced Reka Core - a multimodal LLM that is setting new benchmarks in the field 👀

Designed to process and understand a wide array of inputs including:
- text,
- images,
- audio,
- video

What's even more interesting - in recent evaluations Reka Core has

- outshone Claude 3 Opus in multimodal human evaluations,
- surpassed Gemini Ultra in video understanding,
- demonstrated competitive prowess with GPT-4 in image comprehension.

This positions Reka Core as one of the most capable LLMs available today 👀

Mulitmodal Grok 1.5V

Grok after update can process a wide variety of visual information - including documents, diagrams, charts, screenshots, and photographs.

Grok-1.5V will be available soon to our early testers and existing Grok users.

Krea is introducing Video AI

Ultrarealistic Face Image to Video tool

Microsoft Research just released VASA-1 model, which can generate ultra-realistic talkative video with lip-syncing only from images.

See how insane it is:

Last powerful AI Tools i found

  1. Kaggle

    • Kaggle is the world’s largest data science community. It offers a vast repository of community-edited machine learning models, datasets, and code.

    • As a data scientist or machine learning enthusiast, you can participate in competitions, collaborate on projects, and access valuable resources.

    • Kaggle

  2. Noislee

    • Noislee turns sounds and melodies into music videos.

    • Create captivating music videos by transforming audio content into visual experiences.

    • noisee.ai

  3. Madgicx

    • Madgicx is an AI-powered ads marketing platform.

    • Optimize your advertising campaigns using AI algorithms for better targeting and performance.

    • Madgicx

  4. Fotoram

    Fotoram is a versatile photo editor and collage maker.

    • Enhance photos with basic tools (resize, crop, rotate) or explore advanced features like filters, effects, frames, and text.

    • Fotoram.io

  5. PixiteAI

    • PixelMe! is a generative AI platform for fashion and e-commerce stores.

    • Use AI-generated content for marketing, especially in the fashion industry.

    • https://pixite.ai/

  6. Udio.

    • Udio is very powerful AI music generation app.

    • Create personalized music compositions using AI algorithms.

    • https://www.udio.com/

  7. BasedLabs.

    • BasedLabs. offers text-to-video and image art platforms.

    • Convert text-based content into engaging videos and images.

    • https://www.basedlabs.ai/