Lightning-fast chats with Groq, Pika Lipsync, EMO & more

Check out mind-blowing examples of EMO's lipsync capability

Heather Cooper

Mar 04, 2024

Total reading time around 4 minutes.

Welcome to Visually AI!

There were several interesting developments in generative AI news over the past week.

🔮AI News This Week

Groq

Groq - not to be confused with 𝕏’s Grok - is a new, extremely fast chatbot you can use for free. (link)

Groq is a fast and free chatbot that provides detailed responses in everyday language, making it a valuable tool for generating content ideas.

As someone who relies on chatbots for at least 50% of my written content due to time constraints, I found Groq's speed and level of detail impressive.

While I don't use AI-generated content verbatim, Groq's responses can save time and effort in rewriting and adapting the content to my voice and tone.

Let me show you an example of how you could use Groq.

I asked Groq this question using the Mixtral 8x7B-32k model:

“Please write a short description of Groq, detailing why it is useful for content creation in a neutral tone, in three to four sentences. Please include a concise bulleted list of the features and capabilities that differentiate it from other chatbots.”

This is the response I got in less than 1 second (0.97 seconds):

Groq is a cutting-edge AI tool designed for content creation, offering impressive capabilities for image analysis and text prompt generation. It utilizes advanced computer vision and machine learning techniques to extract relevant features, classify content, and analyze style in images. With its powerful textual description generation and user feedback incorporation, Groq stands out from other chatbots by providing highly accurate and customizable text prompts for users.

Features and capabilities that differentiate Groq from other chatbots include:

• Advanced image preprocessing and feature extraction using Convolutional Neural Networks (CNNs)

• Accurate image classification and style analysis

• High-quality textual description generation tailored for content creation

• User feedback incorporation for iterative improvement

• Compatibility with various image editing tools and generative models for seamless content creation

Here’s a quick visual example of Groq’s speed, with no editing:

Stable Video

Sora videos continue to appear on 𝕏, Instagram, and TikTok - but no word on when Sora will be available for users.

Stability AI launched the Stable Video web app powered by Stable Video 1.1. Stability AI’s Emad Mostaque posted this on 𝕏:

Stable Video generates video from text prompts or an uploaded image. There are limited camera motion options, including Pan, Zoom, and Orbit.

These are videos I downloaded from the Stable Video web app:

Stable Video is available for free during the closed beta. Credits are refreshed daily and each video uses 10 - 11 credits, depending on image-to-video or text-to-video.

You can try it here.

Pika Lip Sync

Pika's Lip Sync feature was unexpected - upload your image, use text-to-voice, or upload an audio file and Pika takes care of the rest.

Pika uses ElevenLabs in the app for text-to-voice generation, but I found it easier to generate a short clip on ElevenLabs.

Early access is available for Pro users. You can try it here.

I tested the feature on my Midjourney image below:

EMO

Pika’s Lipsync was quickly overshadowed by examples of EMO, an “expressive audio-driven portrait-video generation framework.”

Researchers from the Alibaba Group propose that EMO can generate videos of people speaking and singing with facial expressions and body movements to match the input audio for lengths up to the input audio length.

EMO is not available to the public and there has been no release date announced.

Here’s a quick summary of EMO, I recorded from the research page here:

Our T2 Remake AI Film

Our T2 Remake is a feature-length parody film remake of Terminator 2 using AI.

I participated in this collaboration involving more than 50 content creators and produced by Nem Perez and Sway Molina.

I made movie posters for the film and they will be displayed at the premiere at the Nuart Theater in Los Angeles on March 9.

You can purchase tickets and get more details about streaming and the project at t2remake.com.

Here’s one of my posters to be displayed at the premiere (this began as my Midjourney image!):

You could have your AI service, tool, or event seen by Visually AI’s community of over 8,900 subscribers:

Advertise with me

🚀 This Week’s AI Tools

Mistral Le Chat: Versatile AI assistant from the builders of Mistral. (link)

Text-to-Image with SDXL-Lightning: Demo created by AngryPenguin on Hugging Face to utilize the SDXL-Lightning model by ByteDance. (link)

Lexica Art: text-to-image tool released an updated Aperture v4 model that generates nearly 4k resolution images. (link)

Ideogram: AI text-to-image generator updated to Ideogram 1.0 with improved text rendering, prompt adherence & understanding longer, complex prompts, and a prompt enhancement tool called Magic Prompt. (link)

Remix: The mobile text-to-image app added 4 new models, including 3 photorealistic and 1 new anime model, plus the ‘auto-roll’ feature to automatically generate up to 10 images with the push of a button. (link)

Veed.io MP4 Compressor: Free video compression for those large MP4 files you can’t upload due to size limits. (link)

Video Background Removal Tool: Easy and free background removal on HuggingFace. (link)

Everypixel Labs: Pretrained models for lipsync, image captioning, and text-to-speech, plus dev tools for image keywording, UGC, and stock photography scoring. (link)