Ben F PRO

BFFree

labgrab

AI & ML interests

I use my hand drawn sketches as the basis for digital art. I started by using ControlNet Lineart and was so impressed with how specific the inclusion of each drawn element is. I also like to generate img to 3D experiments with sketches, 2D renders and various personal images.

Recent Activity

posted an update 9 days ago

This sketch represents the idea of falling asleep and the mind fragmenting to release the thoughts and bits of the day. Used the https://huggingface.co/microsoft mageflow to give it depth and color. Very accurate detection of the ink drawings small marking and application of interesting details. https://huggingface.co/spaces/microsoft/mage-flow

reacted to unmodeled-tyler's post with 🔥 4 months ago

RESULTS ARE IN! - Videos of each evaluation: https://www.youtube.com/playlist?list=PLkDBfeR-zsShiZ2HpcscFDH-36uDwsl5W - Link to repo: https://github.com/unmodeled-tyler/vessel-browser - https://huggingface.co/quanta-intellect Finally just wrapped up a comparative analysis of my new open source AI browser, Vessel, against Claude Chrome from Anthropic. The test evaluates both web navigation harnesses for speed and efficiency on a simple real-world e-commerce task. Opus 4.6 was used for each of the 3 evaluations, and the results show that Opus 4.6 was AT LEAST 2X FASTER when using Vessel Browser for web navigation in place of Claude Chrome. Results (in order, fastest to slowest) 1. Claude Code + Vessel Browser: 3 minutes and 10s 2. Hermes Agent + Vessel Browser: 4 minutes and 13s 3. Claude Code + Claude Chrome: 7 minutes and 57s Vessel Browser is open source, designed explicitly for agents from the ground-up (it is not a fork of a human browser with AI features bolted on), and supports a local MCP server for agent control, or BYOK custom OAI endpoints. Check it out for yourself!

liked a Space 5 months ago

linoyts/LTX-2-3-First-Last-Frame

View all activity

Organizations

None yet

posted an update 9 days ago

Post

101

This sketch represents the idea of falling asleep and the mind fragmenting to release the thoughts and bits of the day. Used the

microsoft mageflow to give it depth and color. Very accurate detection of the ink drawings small marking and application of interesting details. microsoft/mage-flow

reacted to unmodeled-tyler's post with 🔥 4 months ago

Post

2298

RESULTS ARE IN!

- Videos of each evaluation: https://www.youtube.com/playlist?list=PLkDBfeR-zsShiZ2HpcscFDH-36uDwsl5W
- Link to repo: https://github.com/unmodeled-tyler/vessel-browser
-

quanta-intellect

Finally just wrapped up a comparative analysis of my new open source AI browser, Vessel, against Claude Chrome from Anthropic.

The test evaluates both web navigation harnesses for speed and efficiency on a simple real-world e-commerce task. Opus 4.6 was used for each of the 3 evaluations, and the results show that Opus 4.6 was AT LEAST 2X FASTER when using Vessel Browser for web navigation in place of Claude Chrome.

Results (in order, fastest to slowest)

1. Claude Code + Vessel Browser: 3 minutes and 10s

2. Hermes Agent + Vessel Browser: 4 minutes and 13s

3. Claude Code + Claude Chrome: 7 minutes and 57s

Vessel Browser is open source, designed explicitly for agents from the ground-up (it is not a fork of a human browser with AI features bolted on), and supports a local MCP server for agent control, or BYOK custom OAI endpoints. Check it out for yourself!

liked 2 Spaces 5 months ago

LTX 2.3 First-Last Frame

🎥

153

Generate videos from text, images, and audio

Kokoro TTS

❤

3.42k

Upgraded to v1.0!

liked a Space 6 months ago

KaniTTS2-pt

😻

Generate expressive speech from text with custom voice

posted an update 7 months ago

Post

315

Quietly enjoying my sketch to AI render process. Haven't posted in so long I felt inclined!

These each come from first creating an ink sketch and are processed using ControlNet Lineart setting with a text prompt and sometimes the Flux Style shaping as well.

ENJOY!

liked a Space 11 months ago

MegaTTS 3 Voice Cloning

🎤

300

MegaTTS 3 but with voice cloning!

liked a Space 12 months ago

Higgs Audio Demo

🎤

404

Higgs Audio Demo

upvoted an article about 1 year ago

Article

Welcome GPT OSS, the new open-source model family from OpenAI!

reach-vb, pcuenq, lewtun, clem, Rocketknight1, clefourrier, celinah, Wauplin, marcsun13, pagezyhf, ahadnagy, joaogante

•

Aug 5, 2025

• 514

liked a Space about 1 year ago

VectorFlow

🚀

188

AI Convert Image to .SVG & .AI

upvoted an article about 1 year ago

Article

Transformers Are Getting Old: Variants and Alternatives Exist!

ProCreations

•

Jul 5, 2025

• 46

liked a Space about 1 year ago

Gemma 3n E4B It

⚡

144

Chat with an AI that understands text, images, audio, and video

posted an update about 1 year ago

Post

278

In my painting and drawing practice I had wanted abstraction to be an engagement of the viewer. Giving them an opportunity to add meaning to the work through their own enjoyment and contemplation. Overwhelmingly people would prefer to hear about my intent and then ask me questions about my practice.

I found that to be disappointing because I wanted to them to at least ask questions about the piece. Maybe I wasn't good at giving them a way to connect to the art and therefore they would default to asking 'how often I paint, when did I start painting, how did I find the time to paint" etc. Back in the studio these interactions convinced me that I was on my own pursuit to stay creatively fit and active, but not the kind of artist people would be curious about.

These days I have noticed that when I show people my AI creations or the sketches behind them they ask me about "the ethics of AI, if art is in danger and what I intended to do with them."

The other observation I have had is adults typically flip through the creations quickly. Kids look at a few, ask as few questions about what they are, and then tell me about their own art and things they make with friends. Gotta love kids! Here are a few recent ideas.

liked a Space about 1 year ago

GLM-4.1V-9B-Thinking-API-Demo

🚀

THUDM/GLM-4.1V-9B-Thinking Demo

posted an update about 1 year ago

Post

3113

Working on some chess set concepts. I went towards minimal sculpted shapes then returned to some traditionalism.

posted an update about 1 year ago

Post

731

Love the subtle retro style of this sculptural piece! Took a bit of fine tuning the style but one of the more satisfying things I have created recently.

liked a Space about 1 year ago

Self Forcing Wan 2.1

🎥

326

Real-time video generation

reacted to clem's post with 🧠 about 1 year ago

Post

8124

Today, we're unveiling two new open-source AI robots! HopeJR for $3,000 & Reachy Mini for $300 🤖🤖🤖

Let's go open-source AI robotics!

7 replies

posted an update about 1 year ago

Post

1785

I am a shy artist. Primarily because I don't get motivation from sharing art publicly. I see so much new art daily online that once I begin thinking about where I fit in the mental fatigue becomes counter productive for me.

Recently I shared an album of hundreds of creations with a friend (and singular art fan) and he asked some questions that I felt were interesting enough to create this post on my process and what it teaches me vs what I am seeking.

Specifically I have learned to take ink drawings and create renderings that reveal my actual intention. My digital art goal is to recreate natural details into characters and landscapes that are imagined and deal with my affection for abstraction, deconstruction and humor.

My drawing goals are to be humorous and crafty about how things can be rendered just slightly incorrect to make the viewer see something familiar and recognizable even when its nonsense.

My process is using hysts/ControlNet-v1-1 with Lineart, 50 steps, 14 guidance scale and I give minimal descriptions that are often plain. Example "Really real old dog, plant, and another old dog, with an alligator turtle, posing for a photography portrait".

In the past few months I started taking the ControlNet render to multimodalart/flux-style-shaping and mashing up styles. Here I used a portrait of a Tortise and a dog laying next to each other on a reflective tile floor.

Last night, I took the Flux output and had it described using WillemVH/Image_To_Text_Description which was very accurate given the image.

I then fed the prompt back into Alpha-VLLM/Lumina-Image-2.0

The last step confirmed why I prefer using sketches to language. One, I am a visual artist therefore I have much better nuance with the drawings than with words. Two, my minds eye looks for the distorted. Three MOR FUN.