Claude 3, Pika Sound Effects & ElevenLabs text-to-sound

Plus tons of cool AI tools

Heather Cooper

Mar 11, 2024

Total reading time around 5 minutes.

Welcome to Visually AI!

🔮AI News This Week

Claude 3

Anthropic released the new Claude 3 model family, including Sonnet and Opus, with a third model, Haiku, coming soon.

Takeaways from Anthropic’s announcement about Claude 3:

Claude 3 Opus is Anthropic’s most intelligent model, outperforming peers on benchmarks like expert knowledge (MMLU), reasoning (GPQA), and math (GSM8K). It exhibits near-human comprehension with complex tasks.
The models have increased capabilities in analysis, content creation, code generation, and non-English languages. They also have sophisticated vision capabilities to process photos, charts, graphs, etc.
The models make fewer unnecessary refusals compared to previous versions, showing a more nuanced understanding. Opus demonstrates 2x improvement in accuracy while reducing incorrect answers.
The models will initially offer a 200K context window but are capable of exceeding 1 million tokens. Opus achieved over 99% accuracy on a recall benchmark.
Anthropic has taken steps to make the models trustworthy, mitigating risks around misinformation, privacy, and bias. The models remain at AI Safety Level 2 currently.
Opus and Sonnet are now available in Anthropic's API and on claude.ai. Haiku will be available soon. The models are also accessible through Amazon Bedrock and Google Cloud's Vertex AI.
Anthropic plans further updates to the model family and new features like tool use, interactive coding, and advanced agent capabilities, while continuing to prioritize AI safety.

You can try Claude 3 here.

Apps with multiple chatbot models

Some people prefer to switch between different chatbots easily, and there are several platforms and apps to use:

Perplexity

Perplexity has more multi-modal capabilities with the addition of Claude 3 for chats and Playground v2.5 for image generation.

I use Perplexity frequently and I love the visual results with images - I shared a demonstration in a January article here.

You can upload files in plain text, code, or PDF format (max: 25MB / file). Users on Perplexity’s free plan can upload up to 3 files a day and Pro users get unlimited uploads.

Pro users can upload images (max: 25MB / image) on Perplexity’s mobile or desktop apps. GPT-4 performs the image analysis

AI models: Default, Experimental (Perplexity’s “fast & versatile model”), GPT-4, Claude 3 (latest model), Mistral Large (latest model)

Image generation: Pro users can generate up to 50 images daily with Playground v2.5, DALL•E 3, Stable Diffusion XL

🖼️AI Visual Tools & Tech Updates

AI video tools and new technology are rapidly improving and it feels like 2024 will be a huge year for generative AI in filmmaking and video creation.

Why should you care?

You can use animation / videos for marketing, education, healthcare (patient education through show and tell), gaming, business, entertainment, and much more.

We’re not limited to simply animating images or generating quick clips - now we can add voices and sound effects from a short text prompt.

Pika Sound Effects

Now you can add sound to your video while you generate on Pika 1.0.

The Sound Effects feature can automatically add sound, or you can describe the sound you want and it will generate 3 results to choose from.

It’s currently only available to Pro users and Super Collaborators but will be open to everyone soon. (link)

Here’s a quick demo of the feature:

This is a short collection of some of my first results from my Midjourney images-to-video:

ElevenLabs Sound

ElevenLabs launched Sounds, which lets you generate sound effects with a text prompt.

You get 5 results with each prompt and you can preview and download the ones you want. It’s open for users who joined the waitlist and will be available to everyone soon.

You can try ElevenLabs here.

Here’s a look at some of my first results with Sounds in ElevenLabs:

Kaiber AI

Kaiber’s Motion 3.0 Beta is live and includes smoother motion, photorealistic results, and adjustable audio reactivity for the motion to match the music tempo. (link)

Here’s an example I generated using my uploaded image and music:

Leonardo AI

Connect the AI image generation platform to 1,500+ apps with no installation required, using Pabbly Connect.

This Zapier alternative connects multiple applications to send data back and forth, allowing you to set up automated workflows with marketing, sales, CRM, and many more. You can check it out here.

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

Pallazzo: Interior design app co-founded by tennis star, Venus Williams, has a style quiz to find your “personal design DNA,” with an AI-powered design assistant to help you easily apply your unique style to any room with a few words. (link)

Tracksy: AI-music generator platform added a convenient Text To Music feature that generates original, royalty-free beats in under 1 minute from a short text prompt. (link)

Tonkean: AI-powered no-code platform automates enterprise processes, such as procurement and legal workflows, to increase efficiency, reduce costs, and ensure compliance. (link)

Vital: Personalized meditation guide, and private, audio-guided support sessions generated based on your text prompt describing your feelings or choose one of the 100+ text prompts. (link)

BOOLV: Multi-purpose social media content creation platform powered by OpenAI includes: Object Eraser, Video Enhancer, Background Remover & Blur, and Paraphraser with batch processing. (link)

Gandr.io: Free online collage generator with unlimited photos, customized layouts, text, background colors, and borders. (link)

🖼️ Image Prompts

Prompt: Geometric Mirage in the Desert - reflective dunes with monochrome stripes, mirage of an oasis with colorful abstract palm trees, blue and pink sky gradients