Sudharshan Profile picture
Tweets on Stable Diffusion, LLMs and Gen AI ✍️ ✦ I run an AI product studio ✦ ML engineer. Ex-AI Founder - since 2017. @join_ef @hyperverge @WPI_Robotics
Mar 21, 2023 9 tweets 3 min read
Adobe just launched the beta of Firefly, their AI image creator.

And it's incredible!

- text-to-image
- AI 3D modelling
- AI Video editing

This is the first AI image generator built towards solving creator problems.

Here's a breakdown. "Generative AI made for creators."

1) Image editing + Outpainting

Replace subjects and edit anything
Mar 20, 2023 6 tweets 2 min read
Made a Batman movie in 30 minutes with AI - "The Killing Joke"

Upscaled to 4K and smoothened to 60 fps with AI tools.

Used modelscope text-to-video.

Had an absolute blast - watch till the end for a surprise! How I did this

1. Created 6 clips with modelscope text-to-video. This is a new text-to-video AI

huggingface.co/damo-vilab/mod…
Mar 20, 2023 8 tweets 4 min read
BREAKING 🚨

@runwayml launches Gen-2!

An incredible text-to-video model that's the best one yet.

Here's a breakdown with some examples 🧵 @runwayml "It's like filming something new, without filming anything at all"

Gen-2 offers several modes.

First is Text To Video.

1/ "The late afternoon sun peeking through the window of a New York City loft."
Mar 16, 2023 8 tweets 3 min read
BREAKING 🚨

Microsoft announces Copilot for Microsoft 365.

You can use powerful LLMs in Excel, Word, Powerpoint and their entire suite of apps.

Here's a breakdown from their live event that happened today. 1) Copilot in Word

Copilot in Word writes, edits, summarizes, and creates right alongside you.

"Make the third paragraph more concise. Change the tone of the document to be more casual."
Mar 15, 2023 9 tweets 4 min read
Here's how I gave GPT-4 a photo of a refrigerator and asked it to come up with food recipes in under 60 seconds. ImageImage 1) The multimodal models aren't live yet, so I hacked and used image models from the next best thing - Visual-ChatGPT

Why Visual-ChatGPT?
- Has good Visual Foundation Models
- GPT-4 has been in the works for the past 6 months. Have a very strong hunch that it uses these models.
Mar 15, 2023 13 tweets 5 min read
All the incredible GPT-4 use cases in the last 12 hours.

A thread 🧵 GPT-4 can convert napkins sketches to code. Reminds me of @uizard's pix2code!

Mar 14, 2023 8 tweets 3 min read
BREAKING 🚨

OpenAI just launched GPT-4.

And it’s incredible. Image Read full announcement here : openai.com/research/gpt-4
Mar 8, 2023 4 tweets 2 min read
This is incredible.

Stable Diffusion combined with traditional VFX can generate this 👇 Here's my breakdown and how you can do this too.

1. Pick an animal you like
2. Create a paranormal version of this with img2img + ControlNet.
3. Pass this through EBSynth to transform the video
4. Postprocess with @topazlabs
Mar 7, 2023 6 tweets 2 min read
Google has launched a new AI model, and it's incredible.

Introducing PaLM-E, a multimodal language model across robotics, vision and text.

This is a pretty massive 562B model and beats all previous ones! APPLICATIONS

1/ Robotics

Instruct a robot to "bring me the rice chips from the drawer". Includes multiple planning steps as well as incorporating visual feedback from the robot's camera.
Mar 7, 2023 4 tweets 2 min read
Anyone can download LLaMA weights with this script 😂 Image Fyi, LLaMA is the new LLM from Meta with restricted access. You have to fill up a form to get the weights to run the model.

I can empathize with the intentions behind this decision, but it did face backlash.
Mar 2, 2023 6 tweets 2 min read
Products already using OpenAI's ChatGPT API 👇 1. Snapchat

Introduced My AI, which is basically a friendly ChatGPT inside Snapchat that can provide recommendations and even write Haikus
Feb 20, 2023 7 tweets 3 min read
AI-assisted Coding has exploded in the past 6 months!

These stats from Github's latest report shows the benefits of Copilot.

Here are 5 other AI tools for Coding to improve your productivity👇 Image 1. Codeium (@codeiumdev)

codeium.com
Feb 19, 2023 7 tweets 3 min read
ChatGPT + Wolfram Alpha = 🔥

Check out this trending @huggingface space that makes ChatGPT a Math supercomputer and answers questions like:

→ "What's 2^30?"
→ "How much did it rain in SF today?"
→ "How many calories are there in a cubic light year of ice cream?" HOW DOES IT WORK

→ Ask a question on the Huggingface demo
→ Question gets sent via @Wolfram_Alpha API and you get the correct solution
→ Feed answer to GPT3.5 and construct an answer
→ Use @LangChainAI to interface with GPT3.5
Jun 4, 2021 14 tweets 2 min read
My first paid product crossed $3000 today since I launched 2 weeks back

Community List Stats:
🤯 78 copies sold
🔥 $3051 in Sales
🙏 230+ followers

Sharing my learnings and failures 👇 🤔 Why did I build this?

• Communities are a great place to promote your product. I was wasting time finding communities and figured others would face the same problem.
• Communities are a trend too and there's always opportunity in trends