- Ben's Bites
- Posts
- Daily Digest: Voice translation
Daily Digest: Voice translation
PLUS: Realistic images from Adobe, Kaggle's AI report.
Sign up | Advertise | Ben’s Bites News
Daily Digest #260
Hello folks, here’s what we have today;
PICKS
Dubbing by Eleven Labs - A new voice translation tool that takes your videos/audio and translates them into 29 different languages while keeping how you sound the same. 🍿Our Summary + tips for builders and creators (also below)
Adobe Firefly gets major upgrades. It can now generate more realistic images, thanks to its new model Firefly Image 2. You can have increased control over style, generate vector graphics and more integrations with existing Adobe apps. 🍿Our Summary (also below)
Related read: Adobe’s Project Stardust is a sneak preview of its next-gen AI photo editing engine.Tesla has plans to build a new facility in Austin, Texas to house its powerful Dojo AI supercomputer. It aims to invest over $1 billion in Dojo by 2024, which will be used for training its own AI models (for self-driving) and potentially selling resources like AWS does.🍿Our Summary (also below)
TOP TOOLS
Typeframes - Instant video creation from your website.
Cube by Common Sense Machines - Turn any input into game-engine-ready 3D assets.
Morph Code Index - AI-native OSS semantic code search engine.
PlayFetch - Add LLM features to your app quickly and painlessly.
adminAI - For AE & SDR's who hate admin. Automate follow-ups, notes and CRM updates.
Call Zen – Conversational AI for customer success.
Superdash - Automate extraction & data entry using AI.
TranslateVideo 2.0 - Translate videos to 75+ languages.
WHO’S HIRING IN AI
Glean - AI-powered workplace search. Read our exclusive profile with the founder.
Perplexity - Revolutionary AI search.
Replit - Build software collaboratively with the power of AI.
Deepgram - World-class language models.
Runway - AI creative tools.
Hugging Face - The AI community building the future.
Anthropic - AI research and products with safety first.
Assembly AI - Build AI apps with voice data.
Cohere - Access advanced LLMs through an API.
Character AI - Super-intelligent chatbots.
NEWS
The large language model operations (LLMOps) market map by CB Insights.
Building a 42-inch E Ink Art Frame.
pgvector vs Pinecone: cost and performance
Kaggle has released its 2023 AI report.
Building a recipe chatbot with Langchain, OpenAI and Supabase Vector.
How to use Midjourney and Relume to build a website.
Fine-tuning Mistral 7B using QLoRA.
Disney’s Loki faces backlash over reported use of generative AI.
Saudi-China collaboration raises concerns about access to AI chips.
How non-tech brands use AI - Fashion.
Mitigating stereotypical biases in text to image generative systems - Research by Runway ML.
Replit's new AI model now available on Hugging Face.
Cleanlab raises $25 million to help solve AI models’ data mess.
AMD to acquire Nod.AI - an AI software startup in effort to catch Nvidia.
Modal is now generally available - Run the code for your ML and data jobs in the cloud.
QUICK BITES
At the Adobe MAX conference, Adobe announced several major updates to Firefly. It can now generate more realistic images, thanks to its new model Firefly Image 2. You can have increased control over style, generate vector graphics and more integrations with existing Adobe apps.
What is going on here?
Adobe Firefly is rapidly evolving with new models and features to empower more creativity through AI.
What does this mean?
There are new text-to-image features in the Firefly web app like Generative Match, Photo Settings, Prompt Suggestions, and sharing integrations that aim to make realizing creative visions faster and easier.
The new Firefly Vector model enables the creation of scalable, editable vector graphics through text prompts in Illustrator. The Firefly Design model generates customizable templates from text in Adobe Express.
Firefly Image 2, their new model is able to generate more diverse pictures with higher photorealism. You should follow Kris on X (twitter) for more about how to get the best out of Firefly.
Why should I care?
These Firefly developments are significant for creators. The new models and features allow the generation of high-quality, customizable images, vectors, and templates more easily. The additional creator control over things like image quality and style matching enables you to realise your vision better.
These can save time starting projects or creating assets. And that’s what Adobe is aiming for—to get the adoption by creators who are the bread and butter of its media tools.
Again, the legal and copyright assurance that Adobe provides is a sigh of relief for creators/designers who want to use AI in their day jobs. Adobe’s promise of compensating stock contributors also gets them the ethics points.
How would you feel if we stopped our 'Quick Bites' summaries? |
QUICK BITES
ElevenLabs launches Dubbing - a new voice translation tool that takes your videos/audio and translates them into different languages while keeping how you sound the same.
What is going on here?
ElevenLabs' new AI Dubbing tool can automatically translate speech into 29 languages while maintaining the speaker's original voice and speech patterns.
What does this mean?
Dubbing uses ElevenLabs' proprietary research in areas like multilingual speech synthesis and voice cloning to replicate a speaker's vocal identity and style of delivery when translating their speech. This helps preserve the emotion, nuance, and identity of the original performance in the new language.
Creators can easily dub their content into multiple languages in their original voice, although there’s no lip sync. From quick tests, we also feel translations from English work better than compared to translations to English.
You can try uploading a sample audio/video or a URL to get a 1 minute dubbed result. Tip: Clear site cache and restart the browser to test multiple times.
Why should I care?
Voice dubbing for existing content is a big market. Spotify recently launched its Pilot program with OpenAI to dub podcasts. Spotify’s product is still not open for everyone—while anyone can use the Eleven Labs dubbing feature right now.
If you’re an indie maker who was building for the same problem with Eleven Labs in the backend, don’t stop. It’s easy to put off the project but the interface by Eleven Labs is geared toward general use cases. There is an opportunity to target niche use cases by solving edge cases (like lip sync).
The dubbing isn’t error-free, but if you're a creator, start experimenting now. Compare how your global audience reacts to AI translations compared to subtitles. This is the worst it’s going to be.
QUICK BITES
Tesla has plans to build a new facility in Austin, Texas to house its powerful Dojo AI supercomputer. It aims to invest over $1 billion in Dojo by 2024, which will be used for training its own AI models (for self-driving) and potentially selling resources like AWS does.
What is going on here?
Tesla is investing over $1 billion to develop its own powerful AI supercomputer called Dojo to boost its self-driving capabilities.
What does this mean?
Dojo is supposed to grow into 100 ExaPods (a custom unit used by Tesla) by next year. The existing facility at Palo Alto can accommodate only 7 ExaPods, which is driving the effort behind this new facility. The new Austin facility will supplement its California data centre to accommodate Dojo's growth.
Tesla designed its own specialized AI chip called D1 to run Dojo. Alongside the new facility, it has also doubled the order for these D1 chips from Taiwan.
Why should I care?
Musk, investors and market analysts are betting big on the Dojo supercomputer by Tesla. The rationale behind the backing is twofold:
The access to independent compute resources for better self-driving software
And the option to sell the access to Dojo in an AWS-like manner meaning revenue outside the Tesla vehicle sales.
More broadly, Tesla developing its own AI supercomputer highlights the growing importance of AI across industries. Companies like Tesla now view robust AI systems as a strategic priority.
Unclassifieds - short, sponsored links
VectorAdmin - Manage vector databases at scale with built-in tools that save money and time. Supports Pinecone, Chroma, Weaviate and more.
Ben’s Bites Insights
We have 2 databases that are updated daily which you can access by sharing Ben’s Bites using the link below;
All 10k+ links we’ve covered, easily filterable (1 referral)
6k+ AI company funding rounds from Jan 2022, including investors, amounts, stage etc (3 referrals)
Reply