Ash

2.4K posts

Ash banner
Ash

Ash

@asheem01

▒░|:. Asheem Gujadhur .:|░▒ Original work. I post about my AI experiences. I create my own synthetic selfies using my face. Goal: Realistic selfie on AI body.

127.0.0.1 Katılım Kasım 2015
213 Takip Edilen258 Takipçiler
Sabitlenmiş Tweet
Ash
Ash@asheem01·
Google Veo 3.1 nailed this oversized romantic Paris sunset video. It is now my phone's new lockscreen. #veo3 #veo31 Process: Text to video, 1080p 8s. Platform: @freepik (prompt published there)
English
0
0
2
824
Ash
Ash@asheem01·
Finding a good Ai image upscaler is hard. Below i cropped to her face so you can zoom on the eyelashes & iris. From left to right: - Nano Banana 1 (1K) - Original - ImageUpscaler.com (IU), 4x Up - Upscayl.org @upscayl, 4x Up
Ash tweet media
English
0
0
1
41
Ash
Ash@asheem01·
@itsPolloAI Hiya! Image :: Image Upscaler :: Topaz Labs It doesn't work. Says "Error with third-party service. Credit refunded."
English
0
0
0
4
Pollo AI
Pollo AI@itsPolloAI·
AI showed us what Van Gogh actually saw in his mind. Made with @itsPolloAI Image 2.0
Pollo AI tweet mediaPollo AI tweet mediaPollo AI tweet media
English
15
4
49
5.5K
Ash
Ash@asheem01·
I created a new selfie. Tested it using the same prompt on Nano Banana 1 and Nano Banana 2. Results below. Even if #NanoBanana2 is better at face consistency, i don't like the Image Edits one bit. Synthetic & fat? #NanoBanana 1.0 creations are breathtaking. 🥵
Ash tweet mediaAsh tweet mediaAsh tweet media
English
0
0
1
142
Ash
Ash@asheem01·
A workaround for "Ai HD" not working for images in #Capcut App is to write an upscaling prompt (see Ai Lab screenshot). Here is the big issue in @CapcutApp: - Input img: 768x1344 (2.07MB - Nano Banana 1K png). - Output img: 1440x2560 (737 KB - Capcut 2K jpg). 737KB for an upscaled image DEFEATS THE PURPOSE of the task. The final output should be a minimum of 3 MB jpeg. Who's in charge for such bad decisions at Capcut?
Ash tweet media
Ash@asheem01

Very disgusting experience using @CapcutApp @Capcut_CS. #Capcut lets you enhance single photos. Click "Auto Enhance :: Choose Img :: Enhance Quality :: Choose Ai HD. (Other methods = ugly upscaling). Result is it freezes at 99% and fails with "Couldn't process image quality." Now read their useless copy-pasted response for a 1-week old issue. They have no idea if "Ai HD" is processed locally on phone or at their Capcut servers.

English
0
0
1
121
Ash
Ash@asheem01·
Very disgusting experience using @CapcutApp @Capcut_CS. #Capcut lets you enhance single photos. Click "Auto Enhance :: Choose Img :: Enhance Quality :: Choose Ai HD. (Other methods = ugly upscaling). Result is it freezes at 99% and fails with "Couldn't process image quality." Now read their useless copy-pasted response for a 1-week old issue. They have no idea if "Ai HD" is processed locally on phone or at their Capcut servers.
Ash tweet mediaAsh tweet mediaAsh tweet media
English
0
0
1
182
Ash retweetledi
Krea
Krea@krea_ai·
today, Krea 2 goes live to everyone. to celebrate, we're offering unlimited Krea 2 generations to all our subscribers for the entire week. try it for free 👇
English
64
59
650
777.4K
Ash
Ash@asheem01·
#GrokImagine prompt for "Come here big boy." You eventually get there but very slowly. Enhanced using #Capcut "AI HD" and "Optical Flow." Grok Imagine I2V prompt: "Her gesture is performed as follows: (1) Her hand is held with the palm facing upward (toward the body). (2) The index (fore) finger is extended while the other fingers are clenched inward like a fist. (3) Then the index finger is repeatedly curled inward and extended outward in a hooking motion. Indicating come here and come closer."
English
1
0
4
1K
Ash
Ash@asheem01·
@itsPolloAI In some cases, I found "Seedance 2 Fast" had better prompt understanding than "Seedance 2." I hope people use this PolloAI opportunity to experiment.
English
0
0
2
1.1K
Pollo AI
Pollo AI@itsPolloAI·
Seedance 2.0. Unlimited. May 8–27. 10 days. Unlimited. $0.11 per video.
English
57
23
486
1.6M
Ash
Ash@asheem01·
Why did I use Seedance 2 Fast? Because #Seedance2 kept messing up. Two fails and I gave up.
English
0
0
2
104
Ash
Ash@asheem01·
I was very impressed with those transformer robot clips made using Seedance 2. I always wanted to know if the prompt required was complex? Nokia N810 transformer made in @Capcutapp: - Left: Complex #Seedance2 Fast prompt - Right: Simple let-AI-decide prompt Cpx prompt #Capcut: "Static camera pointing down at a Nokia phone with flipped-out QWERTY keyboard matching the phone in the uploaded image, resting on a check-beige checkered tea towel inside a closed bread container with a curved transparent plastic dome on a kitchen countertop surrounded by soft indoor light shadows. The phone begins a complex mechanical transformation into a miniature bipedal robot, roughly same volume as the Nokia phone. The illuminated glitchy screen brightens and expands upward to form the robot’s head and chest plate, while the display shifts into glowing eye-like structures. The flipped-out keyboard divides and swings outward to become articulated shoulders and arms with mechanical fingers, the main body chassis lifts and separates, and the lower section unfolds downward into powerful metallic legs with visible hydraulic joints and intricate moving parts. Maintaining the phone’s classic Nokia design language with glossy black and silver panels, refined metallic textures, and subtle branding details. The transformation is seamless and highly detailed, with precise moving parts and realistic mechanical motion. The robot looks up curiously at the curved plastic dome of the bread container, and walks on the tea towel causing tiny creases under its footsteps."
English
1
0
2
169
Ash
Ash@asheem01·
Low-light ChatGPT Image 2.0 images look different and have a noticeable grain? Still excellent. I used @Capcutapp's AI Edit on a selfie. #Capcut prompt below. Test: - Left: "surreal" descriptors - Right: Replace/Remove "surreal" with photorealistic #ChatGPTimage #ChatGPTimages2 Use Selfie + Prompt: "A surreal yet elegant dinner scene illuminated by soft, flickering candlelight. The subject — exactly maintaining the same facial features, hairstyle, and body pose as in the reference image — is positioned gracefully within a refined, vintage-inspired setting, without any change in posture or gesture. If the subject appears male, they are dressed in a modern tailored evening suit made of smooth silk or fine satin fabric, exuding understated luxury and contemporary refinement, with soft highlights under candlelight. If the subject appears female, they are dressed in a modern gown crafted from silk or satin, featuring a refined silhouette with elegant skin exposure — shoulders, neckline, or back — radiating sophistication and sensual grace without any metallic shine. The surroundings feature vintage European elements: brass candle holders, fine ceramic dishes, white linen tablecloths, and delicate floral arrangements placed tastefully across the table. Lighting is low and atmospheric, with warm candlelight softly illuminating the subject’s face and creating gentle reflections on glassware and metal surfaces. The overall mood is intimate, mysterious, and romantic — resembling a cinematic still from a luxury fashion magazine editorial, where old-world grace meets contemporary refinement, captured in a surreal, dreamlike tone." File: 1744x2608, 2.5 MB
Ash tweet mediaAsh tweet media
English
0
0
1
641
Ash
Ash@asheem01·
@PrometheanAIX Man, that #GrokImagine! I know there are other video models but i spent way too much time with Grok. (1) Created empty taco shells from your photo (2)(3) Two Grok Imagine draft video. Now we criticise and improve.
Ash tweet media
English
1
0
0
50
Promethean
Promethean@PrometheanAIX·
Taco Tuesday Promptshare #GPTImage2 #NanoBanana #ChatGPT I wanted to nail the idea of fresh tacos as a Taco Tuesday promptshare, with enough art direction to make the ingredients feel like they just hit the plate. That meant building the prompt around texture, steam, falling toppings, matte ceramic surfaces, terracotta tiles, and a 45-degree diner’s perspective that shows the layers instead of flattening the food into a generic overhead shot. The creative direction was really about making the image feel tactile: warm tortillas, seasoned beef, shredded cheese, pico, lettuce, sour cream, lime, and a rustic tabletop that says fresh street-style tacos without turning into a fake restaurant ad. 🔥Prompt: Photorealistic food photography of a plate of steaming fresh street-style beef tacos in soft corn tortillas, generously filled with juicy seasoned beef, shredded lettuce, diced tomatoes, shredded cheddar and Monterey Jack cheese, fresh pico de gallo, and dollops of sour cream. Show the tacos on a rustic matte ceramic plate in an authentic rustic setting with earthy terracotta tile surfaces and warm natural tones. Frame the shot at a 45-degree angle from the diner’s perspective to highlight the layers, texture, and height of the taco fillings. Surround the plate with fresh raw components and garnishes, including lime wedges, cilantro, extra corn tortillas, and small rustic bowls of shredded lettuce, diced tomatoes, pico de gallo, shredded cheese, and sour cream. Add dynamic falling ingredients above the tacos, with bits of cheese, tomato, and lettuce suspended mid-air as if captured with high-speed photography. Include visible steam rising from the tacos to emphasize freshness and warmth. Use soft, appetizing lighting with a shallow depth of field so the tacos stay sharp while the background falls into a soft blur. Matte finishes only, no glossy or reflective surfaces. Rich texture, vibrant fresh ingredients, editorial food styling, premium commercial food photography, highly detailed, realistic, clean composition, square 1:1 format.
Promethean tweet media
Promethean@PrometheanAIX

Food Promptshare: Pizza Food prompts need as much art direction as fashion, product, or film prompts. Maybe even more. Here’s why: #NanoBanana #ChatGPT #Pizza I built the image around the full Margherita language: blistered Neapolitan crust, San Marzano-style tomato red, melted mozzarella white, fresh basil green, and olive oil for that final hit of texture. Then I defined the setting: overhead flat-lay, rustic wood, raw ingredients around the pizza, and a lifted slice with a proper cheese pull. That’s where the food portrayal starts to feel real: ingredients, setting, motion, texture, and mood all working together instead of just asking for “a pizza.” The art direction needs to make people “feel” hungry for this food. You can use this in any image creation platform. 🔥 Prompt: Create a premium food photography image of a freshly baked Neapolitan-style pizza on a rustic wooden table, shot from directly overhead in a flat-lay composition. The pizza should be the main subject, centered in the frame, with a golden-brown puffed crust covered in natural wood-fired char marks. Use simple, fresh Margherita-style toppings: bright red tomato sauce, creamy melted mozzarella, fresh green basil leaves, and small droplets of olive oil. The toppings should feel balanced and naturally arranged, with strong red, white, and green color contrast. Show one slice being lifted from the pizza at the upper edge of the frame, creating a dramatic cheese-pull effect with long strands of melted mozzarella stretching back to the rest of the pizza. The slice should still look connected to the pie, making the image feel appetizing and dynamic. Surround the pizza with fresh raw ingredients to reinforce authenticity and freshness: a bunch of glossy basil leaves, vine-ripened tomatoes, a small bowl of mozzarella, scattered basil leaves, crumbs, and subtle olive oil droplets. Add a high-speed photography feel by freezing a few small ingredients in midair, such as basil leaves, mozzarella pieces, or golden olive oil drops. Use warm, directional natural lighting with deep rustic shadows, crisp texture, and rich detail in the crust, cheese, tomato sauce, basil, and aged wood surface. The overall mood should feel authentic, rustic, Italian, premium, and editorial, like a high-end food magazine cover. Photorealistic, sharp focus, natural food styling, warm tones, rich textures, appetizing composition, no text, no logos, no people.

English
6
1
20
1.4K
Ash
Ash@asheem01·
@PrometheanAIX @capcutapp I am pretty sure they are using prompts that are freely available on X. Perhaps one of yours. 😂
English
0
0
1
12
Ash
Ash@asheem01·
1200 Pro AI credits per month and @capcutapp still puts an AiGC watermark on your content. Shame coz they have fun Selfie Image Edit tools. #Capcut App :: Photo Tools :: AI Photo Editor :: AI Edit - Left: Soft Glow - Right: Magazine Cover - File: 1744x2608, 1.6MB
Ash tweet mediaAsh tweet media
Ash@asheem01

Credit where credit is due @OpenAI. 🥲 At least one good thing happened to me today. @ChatGPTapp did 2 fabulous face transforms. Chatgpt Images 1.5: - Left: My standard M2F face transformation prompt. It beat Nano Banana 1 & 2. - Right: Using Edit function for a slightly rounder & wider chin. GPT gave me 2 options. The correct one is shown below.

English
1
0
2
203
Ash
Ash@asheem01·
@ramizwebti @capcutapp The selfie is from ChatGPT Image 1.5. You can now fully examine how well it did the image edit. x.com/i/status/20371…
Ash@asheem01

Credit where credit is due @OpenAI. 🥲 At least one good thing happened to me today. @ChatGPTapp did 2 fabulous face transforms. Chatgpt Images 1.5: - Left: My standard M2F face transformation prompt. It beat Nano Banana 1 & 2. - Right: Using Edit function for a slightly rounder & wider chin. GPT gave me 2 options. The correct one is shown below.

English
0
0
0
215
Rami
Rami@ramizwebti·
@asheem01 @capcutapp the jump from selfie to vintage film close, up is wild. most people uploading selfies expect enhancement, not full aesthetic reimagining. that gap between input and output is exactly why people are hooked.
English
1
0
1
21
Ash
Ash@asheem01·
I just used ChatGPT Images 2 for the first time. It's an amazing image generator. 🔥 Platform: @CapcutApp #Capcut Model: #ChatgptImages2 Process: Image To Image File details: 2160x3840, 6MB Use of selfie + prompt: "'An intense, high-energy vintage film close-up shot. The camera is placed immediately in front of the windshield, creating a dynamic wide-angle perspective. The character (based on the uploaded image) is driving a retro convertible along a California highway; the car's color is random but stylishly coordinated with the character's vibe. The subject is laughing . The foreground features blurred hands gripping the steering wheel (heavy bokeh), keeping the sharp focus strictly on the facial expression. The aesthetic is high-saturation, strong contrast, sun-drenched, with a distinct retro film grain texture and a 90s editorial look."
Ash tweet mediaAsh tweet mediaAsh tweet media
English
1
0
2
865
Ash
Ash@asheem01·
Comparing Nano Banana Pro against ChatGPT Image 2. Platform: @capcutapp #Capcut Model: #NanoBananaPro Process: Image to Image File: 3072x5504, 3MB Use of selfie + prompt in quote-repost.
Ash tweet mediaAsh tweet mediaAsh tweet media
Ash@asheem01

I just used ChatGPT Images 2 for the first time. It's an amazing image generator. 🔥 Platform: @CapcutApp #Capcut Model: #ChatgptImages2 Process: Image To Image File details: 2160x3840, 6MB Use of selfie + prompt: "'An intense, high-energy vintage film close-up shot. The camera is placed immediately in front of the windshield, creating a dynamic wide-angle perspective. The character (based on the uploaded image) is driving a retro convertible along a California highway; the car's color is random but stylishly coordinated with the character's vibe. The subject is laughing . The foreground features blurred hands gripping the steering wheel (heavy bokeh), keeping the sharp focus strictly on the facial expression. The aesthetic is high-saturation, strong contrast, sun-drenched, with a distinct retro film grain texture and a 90s editorial look."

English
1
0
3
338
Ash
Ash@asheem01·
@Pauldwlon Shocking. It was very quiet & peaceful for the 8 years that I lived there. Old Bull & Bush for a drink. Boris Johnson was the mayor back then.
English
0
0
1
24
Mr.Right
Mr.Right@Pauldwlon·
news.sky.com Have you had enough yet? I have. I’m often in Golders Green, in fact all over London and I’ve been worried about the safety of Jewish people for a long time…places like Stamford Hill really worry me. Enough. Deport. Deport. Deport. #GoldersGreen
English
1
0
0
202
Ash retweetledi
Magnific (formerly Freepik)
Freepik is now Magnific One platform to rewrite the rules of creativity
English
439
382
2.4K
963.7K
Ash
Ash@asheem01·
The @capcutapp #Capcut has 2 ways to upscale a low-quality 400x736 Grok Imagine video. Either "AI UHD" Export, or "Video Quality" option for timeline videos. I am using this 2nd method below: - Left: Enhance Quality :: HD (720p - 4K) - Right: Enhance Quality :: AI HD option Using the normal upscale over-smoothed Pikachu's fur, whereas the AI HD preserved and enhanced the original Pikachu fur texture (as found in the original Grok video).
Ash@asheem01

How can you not love #GrokImagine? A lovely Pikachu visits Ash in his kitchen. Start image: Phone photo Prompt: Click "create your own..." to open Grok Imagine App. Future project: Grok can do it. Now define a slightly different Pikachu.

English
0
0
2
167