The A.I Megathread (LLM , GPT , Development)

bnew · 2024-11-17T04:19:04-0500

1/11
@LangChainAI
We asked, you answered — our State of AI Agents Report is here!

We surveyed 1300+ industry professionals, from developers to business leaders, on how they're using AI agents today — and the results are in.

What are the top use cases for agents? The biggest challenges when building agents? And who's finding success after deploying their agents to production?

Read the full report

LangChain State of AI Agents Report

Here's 5 key insights in the thread below

https://video.twimg.com/ext_tw_video/1857116336973094912/pu/vid/avc1/1218x720/UNMhyAcjOZ95AEuw.mp4

2/11
@LangChainAI
1⃣ Agent adoption is a coin toss, but nearly everyone has plans for it.

About 50% of respondents have agents in production, with mid-sized companies leading the charge. That number is poised to grow, with 78% planning to implement AI agents soon.

3/11
@LangChainAI
2⃣ Research and summarization is the leading agent use case among respondents (at 58%), followed by personal assistance / productivity (54%) and customer service (46%).

AI agents are taking over time-consuming tasks—whether it’s more repetitive tasks for productivity, or handling complex information retrieval and data analysis.

4/11
@LangChainAI
3⃣ Tracing (observability) and human oversight are crucial to keeping agents in check.

Many companies, especially enterprises, favor read-only permissions for their agents and add in extra controls -- including offline evaluation to catch issues before shipping to production.

5/11
@LangChainAI
4⃣ Performance quality is THE major hurdle to deploying agents in production.

Reliability matters — and for many companies, the unpredictability of agent outputs and poor response quality are primary concerns.

6/11
@LangChainAI
5⃣ Agent success stories: The spotlight's on Cursor, Perplexity, and Replit

Companies like @cursor_ai, @perplexity_ai, and @Replit are fan favorites, redefining productivity with AI agents — from code generation to advanced query search.

7/11
@LangChainAI

Curious about the full picture? Check out the full State of AI Agents Report here: LangChain State of AI Agents Report

8/11
@DavidFSWD
what about a discussion about defining "agents"? You just asked a bunch of people if they use "Agents", but there is no qualification for what that means. (first question)

because Agent is a buzzword that doesn't mean anything

9/11
@LangChainAI
Great question! We have a blog discussing the definition of an agent and what it means to be agentic:
What is an AI agent?

10/11
@AIAgentNews
Awesome! Definitely getting featured in this week’s issue

11/11
@nisyron
It's very interesting to not see a discussion about agents for customer support and sales. Is that because there aren't many?

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:21:31-0500

1/3
@HaHoang411

@NexusflowX released Athene-V2 models.
Key points:
- Matches GPT-4 in chat, coding & math
- Good at function calling & agent tasks
- All open-source & available on HuggingFace

/search?q=#LLM /search?q=#Agentic /search?q=#GPT4o

2/3
@HaHoang411
Try chat model here: Nexusflow/Athene-V2-Chat · Hugging Face
Try agent model here:
Nexusflow/Athene-V2-Agent · Hugging Face

3/3
@HaHoang411
They also released a agent benchmark "NexusBench" for testing complex function calling, tool-use and agent capabilities.

GitHub - nexusflowai/NexusBench: Nexusflow function call, tool use, and agent benchmarks.

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:23:26-0500

1/12
@withneo
Introducing NEO: The first Autonomous Machine Learning Engineer.

NEO is a multi-agent system that automates the entire ML workflow, saving engineers thousands of hours of grunt work.

Our recent advancements in multi-step reasoning and memory orchestration have enabled NEO to excel at solving complex Machine Learning problems on its own from data engineering to deployments of ML models.

Benchmarks:
NEO was tested against 50 Kaggle competitions and scored a medal in 26% of them, far superior than the previous state of the art performance of 16.9% by Open AI’s O1 with AIDE scaffolding on the MLE bench

With these capabilities, NEO aims to help every ML engineer become superhuman and focus on real innovation instead of the existing grunt work

Check out NEO in the thread below:

https://video.twimg.com/ext_tw_video/1857448124853157888/pu/vid/avc1/1280x720/aQDSsc6K-IrQC7Q7.mp4

2/12
@withneo
NEO’s score qualifies it as a Kaggle Grandmaster, effectively bringing world-class ML expertise to your fingertips.

Neo is getting ready for early beta users. Join our waitlist here: Neo AI engineer

3/12
@withneo
NEO can perform end to end model building and problem solving.

https://video.twimg.com/ext_tw_video/1857448265312043009/pu/vid/avc1/1280x720/maIhlkt_chBsQxIm.mp4

4/12
@withneo
NEO can manage datasets and perform exploratory analysis by itself for building models.

https://video.twimg.com/ext_tw_video/1857448105379037185/pu/vid/avc1/1112x720/Q64tqh3zHYwt33ix.mp4

5/12
@gneubig
This is cool! Have you evaluated Neo on MLE-Bench? https://arxiv.org/pdf/2410.07095 It'd be interesting to know how it compares to existing solutions like Aide (GitHub - WecoAI/aideml: AIDE: the state-of-the-art machine learning engineer agent, generating machine learning solution code from natural language descriptions.) or OpenHands (GitHub - All-Hands-AI/OpenHands:

OpenHands: Code Less, Make More).

6/12
@withneo
You may find details about our benchmarking here: Neo AI engineer .
We are soon releasing a technical report on the same.

7/12
@josephsemrai
the ml models have started making ml models........

8/12
@withneo
Thanks Joe
Congrats on the launch as well

9/12
@Tharun1850
When is the early access rolling out?

10/12
@withneo
Soon

11/12
@PierrunoYT
Can I try it out?

12/12
@withneo
We will be releasing the early access soon.

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:27:15-0500

1/1
@HaHoang411
Excited to share another breakthrough in AI extraction!

Amazing work by @numind_ai on NuExtract 1.5 smol, powered by @huggingface's SmolLM 1.7B.

Try it here: numind/NuExtract-1.5-smol · Hugging Face

/search?q=#LLM /search?q=#Extract

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:38:59-0500

GitHub - All-Hands-AI/OpenHands: 🙌 OpenHands: Code Less, Make More

🙌 OpenHands: Code Less, Make More. Contribute to All-Hands-AI/OpenHands development by creating an account on GitHub.

github.com

About

OpenHands: Code Less, Make More

all-hands.dev

1/2
@allhands_ai

So far, in November, OpenHands authored or co-authored ~60% of commits to the openhands-resolver GitHub - All-Hands-AI/openhands-resolver: A system that tries to resolve all issues on a github repo with OpenHands. repo and ~20% of the main repo GitHub - All-Hands-AI/OpenHands:

OpenHands: Code Less, Make More

2/2
@allhands_ai
And a bonus - even the script analyzing these stats? Written entirely by OpenHands! See for yourself: All Hands AI

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

1/3
@daniel_mac8
> have you, like me, been on the waiting list for Devin since launch day, anxiously awaiting access?

if yes, why not try OpenHands

(formerly OpenDevin) from @allhands_ai

OpenHands is an open source AI Coding Agent modeled after Devin, but fully open source

Complete with:

> conversational editor
> jupyter notebooks
> web browser

it took me less than 2 mins to get it up and running locally with Docker

next enter your OAI or Anthropic API key and off you go

see in comment for a workflow i've been using recently

https://video.twimg.com/ext_tw_video/1857592338261909504/pu/vid/avc1/1280x720/6s99c20Wn86S6LH5.mp4

2/3
@daniel_mac8

[Quoted tweet]
it took me a while to figure out how to use o1-preview correctly

but when you use it for the right task, it's head and shoulders above other models

here's my AI engineering workflow lately:

1. sketch out a rough idea for software with the help of Sonnet 3.5

2. ask Sonnet 3.5 to create a summary with all the functional and non-functional requirements

3. load that into o1-preview and prompt "create a step-by-step plan broken down into phases for how to implement this using requirements mentioned"

4. use an AI coding agent like OpenHands from @allhands_ai or @Replit agent and feed the plan created by o1-preview to it

with this you can execute on an almost fully automated workflow for software engineering. you'll still have to jump in and fix things, we're not quite to 100% automation yet, but we're getting there

3/3
@daniel_mac8
GitHub - All-Hands-AI/OpenHands:

OpenHands: Code Less, Make More

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:39:13-0500

1/11
@karpathy
Remember exercise pages from textbooks? Large-scale collection of these across all realms of knowledge now moves billions of dollars. Textbooks written primarily for LLMs, compressed to weights, emergent solutions served to humans, or (over time) directly enacted for automation.

2/11
@Emily_Escapor
Are there any more examples? Who is writing these books? Are they already available for a long time?

3/11
@karpathy
data labelers, except the times of just drawing bounding boxes around things are over, now you have to prove a theorem in frontier mathematics and/or critique 5 proofs generated by a state of the art LLM. roughly speaking.

4/11
@rasbt
So, in 2025, textbook writers are basically synonymous with data labelers?

5/11
@mdelbranson
I'm designing my graduate program around synthetic data for math lesson delivery and textbook generation, transcribing every lesson, problem, solution process, and proof for K-12 and undergrad into markdown and releasing it as an OER. I have to defend my proposal this Tuesday.

6/11
@MarkoNjegomir
Some distant future civilization studying Homo Notator:

They invented the press in 1440, only to spend the next 585 years perfecting the art of creating and labeling data so AI could finally be trained in the 21st century.

7/11
@mysticaltech
I ate this stuff for eight years, but can't remember any of it. Now I understand it wasn't meant for humans afterall. At least not most humans

8/11
@teslayoda
Training the trainer.

9/11
@TroyRubert
What a time to be alive!

10/11
@altiamkabir
Fascinating evolution of textbooks into billion-dollar industry.

11/11
@7etsuo
I should write a book.

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

1/3
@wzhao_nlp
Dear @allhands_ai developers: we would like to get OpenHands to do library generation. Would you suggest just kicking off the agent with a prompt "can you complete all function implementations" or explicitly instruct the agent to fill in file by file? @xingyaow_ @gneubig

2/3
@wzhao_nlp
Our current agent wraps aider: commit0/agent/run_agent_no_rich.py at main · commit-0/commit0 where we explicitly instruct aider to fill file by file and run unit tests to perform iterative improvement.

3/3
@wzhao_nlp
our current agent design wraps aider to do each of the following steps. What I am unsure about is if we should just replace aider with openhands, or if you are doing this in an entirely different way?

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:41:01-0500

1/3
@HaHoang411
How this I missed this?
Awesome work by @bria_ai_ team! They've just released RMBG 2.0, their new state-of-the-art background removal model.

Key notes:
- It has only 968M parameters.
- Nearly matches Remove Background from Image for Free – remove.bg's performance (97%) in photorealistic

https://video.twimg.com/ext_tw_video/1857359766181355520/pu/vid/avc1/928x720/9v1gr8wfegwRnj-Y.mp4

2/3
@HaHoang411
Official blog: Bria's New State-of-the-Art Remove Background 2.0 Outperforms the Competition

3/3
@HaHoang411
Give it a shot here: BRIA RMBG 2.0 - a Hugging Face Space by briaai

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:50:37-0500

Bria's New State-of-the-Art Remove Background 2.0 Outperforms the Competition

Discover how Bria’s new RMBG 2.0+ model revolutionizes AI-driven background removal, outperforming leading competitors in accuracy and consistency for diverse real-world applications.

blog.bria.ai

Bria's New State-of-the-Art Remove Background 2.0 Outperforms the Competition

Bria.ai : Nov 12, 2024 9:34:52 AM
Technology

Play

Bria's New State-of-the-Art Remove Background 2.0 Outperforms the Competition

AI-generated audio

4:34

Intro

In the rapidly evolving field of AI-driven image processing, background removal remains one of the most challenging tasks, particularly in workflows that involve complex image compositions.

This advanced open-source model enables development teams to streamline and scale processes with precise and reliable background removal. By integrating RMBG 2.0 into their workflows, teams can automate intricate image processing tasks, facilitating scalable content creation across various industries such as e-commerce, image editing, stock media, gaming, and more.

What’s New in RMBG 2.0

Bria's RMBG 2.0+ sets a new standard in open-source background removal by overcoming the limitations of RMBG 1.4. While the original model was effective, RMBG 2.0+ advances its capabilities to deliver state-of-the-art, industry-leading results that distinguish it from competitors.

RMBG 2.0+ provides unmatched precision in background removal with exceptional accuracy and consistency, even in complex scenes. It is available through a Model, API, and iframe for commercial use, with a commercial agreement required. For non-commercial use, it is only accessible through the model.

Impact on Bria API Routes

Separating and layering a “foreground” from a ”background” is the foundation of many complex AI Tasks. Thus, such an upgrade systematically improves any downstream endpoints utilizing this core function.

watch_display_in_a_luxury_store__background_of_the_store_with_people_undefined (1)

2.3FAST - a_fluffy_white_cat__sitting_on_the_glass_table_with_a_slight_reflection_undernea_Seed1372292595

See How it Compares to the Competition.

See how our new state-of-the-art 2.0 Background Removal capability consistently outperforms competitors like BiRefNe, Photoshop, and Remove.bg, setting a new benchmark for quality and commercial readiness.

We tested these capabilities and measured their outputs for commercial readiness and quality of results.

Three top industry-leading Remove Background, Open Source Models.,

Competitor Overview

[td width="226.5px"]

RMBG 2.0

[/td]
[td width="226.5px"]

[/td]
[td width="226.5px"]

[/td]
[td width="226.5px"]

[/td]

[td width="226.5px"]

It offers Remove Background 2.0, which has an open-source model, API, and iFrame.

[/td][td width="226.5px"]

An open-source model was used, and it was the latest open SOTA

[/td][td width="226.5px"]

Photoshop Background Removal offers an Interface and API

[/td][td width="226.5px"]

Company specializing in Background Removal, which is available via a web app and an API

[/td]

Learn More about Bria’s AI Image Editing Products

Methodology

To evaluate RMBG 2.0+ against other leading background removal models, we benchmarked its performance on diverse images, assessing each model’s output for commercial readiness. Results were rated on a four-point scale: Very Bad for outputs with significant artifacts, Bad for conceptually accurate but unusable results, Good for outputs needing minor adjustments, and Very Good for near-perfect, ready-to-use segmentation.

Our test included diverse subject compositions (e.g., people, objects, animals, text) and photorealistic and non-photorealistic images to gauge versatility. We used images with simple, complex, and transparent backgrounds to test precision and analyzed foreground complexity with single and multiple elements.

This approach gave us a comprehensive view of how RMBG 2.0+ performs across various real-world scenarios, solidifying its position as a state-of-the-art model for consistent, high-quality background removal.

Benchmarking Results

SIZE=0px]1

SIZE=0px]2[/SIZE]
SIZE=0px]3[/SIZE]
SIZE=0px]4[/SIZE]

Models’ Success Rate
Photorealistic vs. Illustrative Backgrounds
Model’s Consistency
Solid vs. Complex Background

This graph illustrates the percentage of benchmarks in which each model achieved usable results, defined as a score of "Good" or "Very Good."

SIZE=16px]BiRefNet: Bria’s new RMBG 2.0 model outperforms the current open-source SOTA, BiRefNet, by a significant margin (90% vs. 85%), positioning RMBG 2.0 as the new state-of-the-art in open-source background removal.[/SIZE]

SIZE=5]Adobe Photoshop: Bria’s RMBG 2.0 surpasses Adobe Photoshop’s background removal capabilities by a large margin (90% vs. 46%).

SIZE=5]Remove.bg: Our model delivers competitive results against leading commercial solutions like remove.bg (90% vs. 97%). Particularly in the photorealistic use case, our primary focus, RMBG 2.0, achieved 92% accuracy compared to the removal.bg’s 97%.[/SIZE]
SIZE=5]RMBG 1.4: The new RMBG 2.0 model dramatically improves upon our previous RMBG 1.4 model (90% vs. 74%). This upgrade addresses critical issues in RMBG 1.4, with marked improvements in consistency, reduced indecisiveness, and significantly fewer misclassifications.[/SIZE]

[/SIZE]

Conclusion

The comprehensive benchmarking of RMBG 2.0+ demonstrates its remarkable advancement in background removal, positioning it as the new state-of-the-art solution among open-source models. RMBG 2.0+ consistently outperforms other models, particularly in handling complex backgrounds, where it excels beyond BiRefNet and Adobe Photoshop. Its high accuracy and reliability, especially in photorealistic scenarios, make it highly competitive with commercial solutions like remove.bg.

RMBG 2.0+ addresses critical issues in earlier models, such as inconsistent output and indecisiveness in complex scenes. This upgrade enhances the precision and robustness of Bria’s background removal capabilities and delivers commercial-grade results suitable for diverse real-world applications. Through its open-source availability, RMBG 2.0+ empowers development teams with a reliable, scalable tool for high-quality background removal, setting a new benchmark in the industry.

[td width="269.828px"]

Browse Our
API Documentation

[/td]

[td width="272.172px"]

Demo Bria’s Remove Background 2.0 Capability in Our Console

[/td]

[td width="269.828px"]
Demo Bria’s Remove Background 2.0 Capability in Hugging Face [/td]
[td width="272.172px"]
Learn more in our Hugging Face model card
[/td]

[/SIZE]

bnew · 2024-11-17T04:52:20-0500

1/3
@HaHoang411

Nexa AI team created OmniVision-968M, the world's smallest vision-language model that packs quite a punch for its size.

Key highlights:
- Built on solid foundations: Qwen2.5-0.5B + SigLIP-400M
- Practical edge deployment: Runs on ~1GB RAM

/search?q=#SmolLM /search?q=#Multimodal

https://video.twimg.com/ext_tw_video/1857355906670800896/pu/vid/avc1/1280x720/1ZEF3mCgMRo6v8BU.mp4

2/3
@HaHoang411
Official annoucement: OmniVision-968M: World's Smallest Vision Language Model

3/3
@JulienBlanchon
Pretty interesting

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:54:01-0500

1/8
@rohanpaul_ai
Open-source alternative to OpenAI's voice.

Extends text-based LLM to have native "listening" ability. Think of it as an open data, open weight, on device Siri.

What it offers:

Implements native "listening" abilities in LLMs, similar to Siri but runs locally on-device

Latest v0.4 achieves 64.63 MMLU score with enhanced context handling and noise management

Provides complete training pipeline combining speech noise and multi-turn conversation data

Includes ready-to-use demos via WebUI, Gradio interface, and Colab notebooks

2/8
@rohanpaul_ai

GitHub - homebrewltd/ichigo: Local realtime voice AI

3/8
@antona23
Thanks for sharing

4/8
@rohanpaul_ai

5/8
@antonpictures
Does it run on an M1 mac with 8gb ? lol I need better hardware. This mixed with llama 11b vision and some RAG, that's a robot

6/8
@gpt_biz
This sounds fantastic! It's great to see an open-source voice assistant that can run locally—definitely worth a try for those who value privacy and customization.

7/8
@techfusionjb
Native listening ability on-device? That's huge!

8/8
@sky20086
save thread

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:54:37-0500

1/1
@HaHoang411

Incredible milestone for open-source AI! @PIeIAs just released Common Corpus - a massive 2T token dataset that's truly game-changing.
Why I'm excited about this: 100% permissively licensed data (finally!)
/search?q=#AI /search?q=#LLM /search?q=#OpensourceDataset
PleIAs/common_corpus · Datasets at Hugging Face

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:55:02-0500

1/4
@HaHoang411
Awesome work by the @deepseek_ai on JanusFlow. They've managed to elegantly combine image understanding and generation in a single model using rectified flow with autoregressive language models.
/search?q=#AI /search?q=#OpenSource /search?q=#Any2Any

https://video.twimg.com/ext_tw_video/1856952968982974464/pu/vid/avc1/1082x720/dRnQFduGCqCMmQdS.mp4

2/4
@HaHoang411
For OCR task it's pretty good but still has problems.

https://video.twimg.com/ext_tw_video/1856961861570199552/pu/vid/avc1/1082x720/QSkGSO5eJt-aZzyO.mp4

3/4
@HaHoang411
They also have some good examples to generate image from text. But to be honest, the image gen needs a lot improvements.

https://video.twimg.com/ext_tw_video/1856962863597813762/pu/vid/avc1/1082x720/_cTXoWRetE0IbV71.mp4

4/4
@HaHoang411
The model: deepseek-ai/JanusFlow-1.3B · Hugging Face
Try here: JanusFlow 1.3B - a Hugging Face Space by deepseek-ai

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:57:15-0500

1/3
@jandotai
Jan v0.5.8 is out: Jan supports Qwen2.5-Coder 14B & 32B through Cortex

Highlights

- A new engine: Jan now run models via @cortex_so
- @Alibaba_Qwen's Coder 14B & 32B Support
- Supports markdown rendering on user messages

and various UI/UX enhancements

Update your product or download the latest.
Jan: Open source ChatGPT-alternative that runs 100% offline - Jan

2/3
@whatmeigel
Would be nice to also have some smaller models available.

3/3
@jandotai
Ah, Jan already supports GGUF models, so you can run smaller models if you prefer.

Just find the model on @huggingface & paste the link into Jan's Hub

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T04:57:47-0500

1/11
@sunomusic
v4 is coming soon

https://video.twimg.com/ext_tw_video/1854955202001907712/pu/vid/avc1/720x1280/JFFDhLm1PGV4YWjU.mp4

2/11
@ThenAMug
I hate how corny most of the Hip-Hop/Rap generations are. Hopefully, this update brings some justice to the genre. @sunomusic

3/11
@sunomusic
we posted a hip hop song made with v4 this morning!

4/11
@Jesseclinton_
People who don’t understand creativity will knock this. True creators don’t just use a prompt; they see AI/Suno as part of a process. Bach with a strings VST and a chord generator—he’d innovate, not dismiss. Creativity evolves with tools, not despite them.

5/11
@sunomusic
said beautifully. thanks!

6/11
@StevenDarlow
@nickfloats wasn’t lying!

Wow, that is a game changer!! (again)

7/11
@sunomusic
thanks steve! excited for you to try it

8/11
@NahPlaya1212
Take everything I own

9/11
@sunomusic

10/11
@cooljellyx
Add in painting, and the game is f*cking over.

11/11
@sunomusic
we have inpainting!

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

1/11
@sunomusic
Burkinabe funk, but make it v4

Thank you to all our alpha testers for your feedback on v4 this past week. With your help, we’re now working hard at adding some finishing touches. Thank you everyone for your excitement and patience as we

https://video.twimg.com/ext_tw_video/1857478323317706752/pu/vid/avc1/720x1280/mOnbevzYhNIAIjDQ.mp4

2/11
@AIandDesign
It's amazing!

3/11
@sunomusic
thanks Marco! really glad to hear that. and appreciate your feedback the last few days!

4/11
@JoeProAI
I did a song in anticipation but felt cliche to post. I do have to say whomever is running this update coming soon a campaign is doing a great job. I'm super stoked for this.

5/11
@sunomusic
thank you

6/11
@SwisherYard
Will it have the option to enhance existing songs?

7/11
@sunomusic
Yes - we will have remastering as an option on your existing songs!

8/11
@lxe
This is absolutely bonkers. I've been addicted to suno.

9/11
@sunomusic
appreciate your support and glad you’ve been able to make your own music with suno!

10/11
@WorldEverett
Can't wait, it sounds great

11/11
@RichSilverX
I want it so bad!!

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

1/11
@sunomusic
v4 is coming soon

https://video.twimg.com/ext_tw_video/1856106054909632512/pu/vid/avc1/720x1280/tEGw361SAgwBxdzT.mp4

2/11
@Mitchnavarro7
I'm so pumped for this!

3/11
@gopro_audio
I got 500 songs to re do

lets go

4/11
@jasonjdxb
Can’t wait for SUNO V4! Can we get an upgrade where we can upload our original songs and SUNO AI creates vocals on the entire song using time stamps/prompts for guidance?

5/11
@Emily_Escapor
I hope soon means next week, and I hope you guys start the initial preparation for V5 and start training before the end of the year on the B200 cluster; we need everyone to make music.
I hope we also significantly improve how to control the instruments and advanced control over vocals.

6/11
@mckaywrigley
Can’t come soon enough

7/11
@blizaine
Wow. Tomorrow then?

8/11
@ibrvndy
How soon we talkin?

9/11
@Faedriel
hypeee

10/11
@david_vipernz
It's already best value for anything ever. And it keeps getting better!

11/11
@Liinad_De_Varge
Already forgot you existed

But going to try out the new version for sure!

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

1/11
@nickfloats
something comin' @sunomusic

https://video.twimg.com/ext_tw_video/1854964192417038336/pu/vid/avc1/720x1278/9RQop-MC3yfc638K.mp4

2/11
@dustinhollywood

3/11
@DMctendo
Put it on hi fi speakers in a club and see why it’s not gonna be doing much of anything… yet

4/11
@dkardonsky_
bullish

5/11
@jaredeasley
Fantastic

6/11
@gpt_biz
Excited to see what’s coming next from @sunomusic looking forward to it

7/11
@edh_wow
Honestly, I love Suno. Has the rapping gotten better?

8/11
@ikamanu
Love it. Can you share the prompt for this one?

9/11
@NoodleNakamoto
I'm NGL, this is an absolute banger. Please @nickfloats - if you can extend this, I'd love to hear more.

10/11
@opensaysmani
V4 on its way - can’t wait

11/11
@be_high
API? Word level timestamps ?

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

1/11
@sunomusic
Make a song from any image or video with Suno Scenes.

: Disco Sculptures by ClaraCo

https://video.twimg.com/ext_tw_video/1852821409162207232/pu/vid/avc1/720x1402/WY2XWQmD_XMSALDE.mp4

2/11
@frownsOfficial

https://video.twimg.com/ext_tw_video/1852822140372062210/pu/vid/avc1/720x720/2ysRvDKZDqQ1MC4A.mp4

3/11
@sunomusic

4/11
@UriGil3
Youre too focused on mainstream music. Wish there was more creativity

5/11
@sunomusic
happy to take style suggestions

6/11
@luckycreative_o
And for Android users?

7/11
@sunomusic
Good news - Suno on Android is now open for pre-registration at Suno - AI Music - Apps on Google Play! Have you signed up?

8/11
@BromfieldDuane
when is this coming for android?

9/11
@sunomusic
Good news - Suno on Android is now open for pre-registration at Suno - AI Music - Apps on Google Play! Have you signed up?

10/11
@Mitchnavarro7
I have an android

11/11
@jessyseonoob

https://video.twimg.com/ext_tw_video/1852943354407100417/pu/vid/avc1/720x948/0kx3YGDa890zieDc.mp4

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

bnew · 2024-11-17T05:05:26-0500

1/7
@homebrewltd

Ichigo is smarter now with v0.4

- MMLU score up to 64.66
- Better at picking out voices in noisy settings
- Tracks complex, multi-turn conversations better

Links:
- GitHub: <GitHub - homebrewltd/ichigo: Local realtime voice AI>
- Live demo: https://ichigo.homebrew.ltd/
- Model weights:

Ichigo v0.4 - a homebrewltd Collection

2/7
@daniel_nguyenx
Nice!!!

3/7
@carsenklock

4/7
@tuanlda78202

5/7
@wangye1644294

6/7
@pokachi2023
Small increment for the community!

7/7
@Sathees89347227

To post tweets in this format, more info here: https://www.thecoli.com/threads/tips-and-tricks-for-posting-the-coli-megathread.984734/post-52211196

The A.I Megathread (LLM , GPT , Development)

Veteran

Veteran

Veteran

Veteran

Veteran

Veteran

Veteran

Veteran

Bria's New State-of-the-Art Remove Background 2.0 Outperforms the Competition​

Intro​

What’s New in RMBG 2.0​

Impact on Bria API Routes​

​

​

​

See How it Compares to the Competition.​

Competitor Overview​

​

Methodology​

Benchmarking Results​

Conclusion​

Veteran

Veteran

Veteran

Veteran

Veteran

Veteran

Veteran

Similar threads

Bria's New State-of-the-Art Remove Background 2.0 Outperforms the Competition

Intro

What’s New in RMBG 2.0

Impact on Bria API Routes

See How it Compares to the Competition.

Competitor Overview

Methodology

Benchmarking Results

Conclusion