Sign in

Doron Adler

@norod78.bsky.social
232 followers 856 following 52 posts

Been fooling around with generative machine learning since before it became mainstream 😊 linktr.ee/Norod78

PostsRepliesMedia
Doron Adler @norod78.bsky.social · 22/05/2025
That's awesome!
010
Doron Adler @norod78.bsky.social · 22/05/2025
Here, I fixed it with AI
040
Doron Adler @norod78.bsky.social · 10/04/2025
כ 3.5, אני בן 46
010
Doron Adler @norod78.bsky.social · 25/03/2025
פריחה ורודה על רקע שמים כחולים זה באמת יפה. או לחלופין אולי למצוא זווית שמדגישה את הניגוד בין היופי של העץ לבין הג׳יפה מסביב, אם יש משהו בולט מספיק שמשלב את שניהם
020
Doron Adler @norod78.bsky.social · 25/03/2025
נשמע כמו פוטנציאל לקומפוזיציה מעניינת
110
Doron Adler @norod78.bsky.social · 19/03/2025
ooooh Nice!
000
Doron Adler @norod78.bsky.social · 10/03/2025
Inserted Princess Peach into a video of my kids (Filmed at Dortmund Zoo this summer) using Pikadditions and then did a vid2vid Style-Transfer with Mago studio AI.
050
Doron Adler @norod78.bsky.social · 10/03/2025
קרה לי בול אותו הדבר (שוב) שבוע שעבר. כל פעם שזה קורה אני אומר לעצמי ״טוב, בפעם הבאה אני אעשה תזכורת ביומן - כמה ימים לפני מועד פקיעת התוקף״ 🫠
030
Reposted by Doron Adler
luokai @luok.ai · 05/03/2025
- Voice conversion: It supports various conversion functions like voice - to - voice, voice - to - instrument, speech - to - singing, and language transformation. Artists can turn their voices into various sounds for unique sound effects.
121
Doron Adler @norod78.bsky.social · 23/02/2025
I actually enjoyed it. The first ~1/4 of the film was kind of meh, but then it become fun and all and all I'm happy I watched it.
000
Doron Adler @norod78.bsky.social · 17/01/2025
It also gets all other summarization types wrong, not just from news sources but I guess that in the case of news it is more concerning
010
Doron Adler @norod78.bsky.social · 13/01/2025
נייס, מסתבר שזה יום ההולדת של הברווזון של אריק www.google.com/search?q=rub...
google.com
rubber duck day - Google Search
020
Doron Adler @norod78.bsky.social · 09/01/2025
Today I learned that you can backup your entire iCloud, Facebook and Instagram media to Google-Photos with just a few clicks at Google One's "Storage" page: one.google.com/storage
030
Doron Adler @norod78.bsky.social · 08/01/2025
A beer bottle robot, dancing hip hop. Image generated with Flux, converted to 3D with Trellis and animated with Mixamo. VideoLeap was used for editing and soundtrack. The 3D model is available for download on Sketchfab: skfb.ly/ptABZ
040
Doron Adler @norod78.bsky.social · 30/12/2024
Love this shave & haircut spell
020
Doron Adler @norod78.bsky.social · 25/12/2024
וגם יש לזה פרצוף חמוד 😊
140
Doron Adler @norod78.bsky.social · 25/12/2024
זה יפה ממש!
130
Reposted by Doron Adler
Jeremy Howard @howard.fm · 19/12/2024
I'll get straight to the point. We trained 2 new models. Like BERT, but modern. ModernBERT. Not some hypey GenAI thing, but a proper workhorse model, for retrieval, classification, etc. Real practical stuff. It's much faster, more accurate, longer context, and more useful. 🧵
19620147
Doron Adler @norod78.bsky.social · 18/12/2024
זה חמוד! אז השתמשתי ב טרליס ומקסימו ועכשיו הוא מתאגרף skfb.ly/psWoU
skfb.ly
Buff Kirby - Boxing (Mixamo) - Download Free 3D model by Norod78 (@Norod)
The Buf-Kirby meme image converted to 3D using Microsoft Trellis and animated with Adobe Mixamo License was set as “CC Attribution-NonCommercial-ShareAlike” because the model contains Mixamo animatio...
110
Doron Adler @norod78.bsky.social · 17/12/2024
Is there a feasibility for a hybrid approach of sorts? Like a "Tool call" to a classic NLP model which output could guide / advise / assist the LLM in providing better results?
100
Doron Adler @norod78.bsky.social · 11/12/2024
FLUX Style Shaping is really cool! huggingface.co/spaces/multi... Note it it will not preserve people's identity, but when you can match a content and style with 'similar overlapping properties', the results can be stunning.
010
Doron Adler @norod78.bsky.social · 09/12/2024
I agree. Its ability to generate a UV mapped mesh with a texture already puts it above (almost?) all other models I've tried so far and in addition, it really is excellent in generating the mesh itself
010
Doron Adler @norod78.bsky.social · 09/12/2024
Flow: Image of cucumbers converted to 3D with huggingface.co/spaces/Jeffr... converted to a .obj/mtl/png trio, attached the arms back to the body with SculptGL stephaneginier.com/sculptgl/ fixed orientation, scaling and reassigned texture with MeshLab. Rig and animation with www.mixamo.com
130
Doron Adler @norod78.bsky.social · 09/12/2024
We live in a future where you can take a picture of a cucumber platter at the office and have it dance a few minutes later 🥒🪩
182
Doron Adler @norod78.bsky.social · 08/12/2024
Nah it's just hiding in a dark corner ^_^
010
Doron Adler @norod78.bsky.social · 08/12/2024
A socially awkward potato turns to the dark side :)
100
Doron Adler @norod78.bsky.social · 08/12/2024
Yeah, I played with it yesterday and it's the best "Image to 3D mesh" I've tried so far skfb.ly/psGEz
skfb.ly
Another socially awkward potato [AI Generated] - Download Free 3D model by Norod78 (@Norod)
Generated from image using Microsoft TRELLIS - Another socially awkward potato [AI Generated] - Download Free 3D model by Norod78 (@Norod)
100
Doron Adler @norod78.bsky.social · 08/12/2024
Yay ^_^
010
Doron Adler @norod78.bsky.social · 08/12/2024
Added some cheap VFX for the atmosphere 😅
171
Doron Adler @norod78.bsky.social · 08/12/2024
By naming them, the contract is sealed.
020
Doron Adler @norod78.bsky.social · 08/12/2024
Four images, but so many emotions 😊
020
Doron Adler @norod78.bsky.social · 07/12/2024
תתחדשי ^_^
000
Doron Adler @norod78.bsky.social · 06/12/2024
XKCD is a reliable source and I don't think I like the other option. (source: xkcd.com/2592/ )
010
Doron Adler @norod78.bsky.social · 04/12/2024
מעניין! כנראה שמרחק של מעל 35 ס״מ מהכיריים זה כבר בטוח לכל סוגי הקוצבים. במרחק קרוב יותר הסכמה היא בעיקר בתנאים מסויימים למי שיש קוצב מסוג unipolar electrode pubmed.ncbi.nlm.nih.gov/16635999/
pubmed.ncbi.nlm.nih.gov
Do induction cooktops interfere with cardiac pacemakers? - PubMed
Patients are at risk if the implant is unipolar and left-sided, if they stand as close as possible to the induction cooktop, and if the pot is not concentric with the induction coil. Unipolar pacing s...
110
Doron Adler @norod78.bsky.social · 02/12/2024
So cool! prompt: "What are the 10 most popular languages based on the values in langs (note that langs is an array which may contain several languages)"
A list of the most popular languages on Bluesky. EN, JA, PT, ES, DE, FR, KO, NL
030
Doron Adler @norod78.bsky.social · 29/11/2024
Well true, but high quality photos can be applied noise to, downsized and resized back, have random characters written on them, any combo of such nature and also be used as high quality source. While potato quality is more limited in this sense :)
000
Doron Adler @norod78.bsky.social · 28/11/2024
כן, ב 2022 יצא SE 3rd generation במקביל לאייפון 13 (וה 13 מיני)
100
Doron Adler @norod78.bsky.social · 28/11/2024
כרגע ה SE הכי עדכני הוא מ 2022 ל SE יש מסך בגודל 4.7 אינצ׳ ויש גם אייפון 13 מיני עם מסך בגודל 5.4 אינצ׳.
010
Doron Adler @norod78.bsky.social · 28/11/2024
אכן. ויש פלוס שזה כמו הרגיל עם מסך גדול יותר ויש פרו מקס שזה כמו הפרו עם מסך גדול יותר.
020
Doron Adler @norod78.bsky.social · 28/11/2024
Thank you very much
000
Doron Adler @norod78.bsky.social · 27/11/2024
Cool! it can 'enumerate' objects seen in an image and output the result as valid JSON while adhering to the instructions in the prompt. Well done! 👏 Output: {'object_list': ['minion', 'lightsaber', 'power strip', 'cable', 'box', 'robot']} Input prompt in ALT, top-p: 0.2, temp: 0.1
Your task is to analyze an image and output a list of all distinct objects, people, animals, etc. that you can identify in the image. For every image the user sends, please study that image carefully and compile a json-like list of every meaningful object or entity you can discern. Each item should be on a separate line. The list should be thorough in capturing all key elements of the image, but also concise, avoiding redundancy. Minor background details can be omitted. Output your list in the following format: {"object_list":[item 1, item 2 , item 3]} The formatting of this list is important, as it is meant to be easily machine-readable for later processing. Focus only on the literal contents of the image. Do not include any subjective analysis, guesses about the context, or details that aren't clearly visible. Provide only the object list, without any other discussion or response. If no object were detected, just return: {"object_list":[]} Each item must appear only once, never repeat the same item.
162
Doron Adler @norod78.bsky.social · 25/11/2024
הלוחמים שעוברים את המבחן מקבלים סיכת חי
010
Doron Adler @norod78.bsky.social · 24/11/2024
גרייט מיינדס ^_^ bsky.app/profile/noro...
010
Doron Adler @norod78.bsky.social · 23/11/2024
הא! 😄
000
Doron Adler @norod78.bsky.social · 23/11/2024
Name a show you’re positive no one remembers but you
020
Doron Adler @norod78.bsky.social · 23/11/2024
Hey hey hey Interbang?! היה שם גמל מדבר, אם אני זוכר נכון
110
Doron Adler @norod78.bsky.social · 21/11/2024
חתול תקול 😺
030
Doron Adler @norod78.bsky.social · 18/11/2024
Nice. I like the "Nanoset" part, was wondering what's a good way of weight-mixing datasets and pre-tokenizing them for training and that looks like a very interesting solution github.com/huggingface/...
github.com
000
Doron Adler @norod78.bsky.social · 18/11/2024
Cool! Is there somewhere a list of the languages which were part of the training set?
000