Award-winning CG artist turned AI researcher. Founder @ DyzenLabs. Building controllable AI tools: Anvi, Basher, Mohmaya. Taste over pure gen.
Bangalore, India
Joined December 2018
- Tweets1.7K
- Following1.2K
- Followers558
- Likes6.1K
Pinned Tweet
Next Gen Typography Tools are going to be soo crazy!! 🤯
Combined Touch type by @schultzschultz_ with LCM + sdxl
@peteromallet
@HelloCivitai
@StableDiffusion
#aianimation #aivideo #stablediffusion #SDXL #typography #LCM
The funniest part about AI video generation is that it can perfectly clone your face, hair, and micro-expressions, just to show you a reality that doesn't exist.
I’ve never growled on a stage or been in a street fight like jackie chan, yet here we are. It’s so realistic it’s eerie, but I feel totally detached from it.
Tell me: Is this cool or low-key terrifying? What would your AI clone be doing right now? Let’s talk 👇
🤖 Made with AI
2
1
180
and its crazy how fast we reached at this point, here is me in 2023 trying to make a video with my likeness, back when i had long hairs:
nitter.gq/mjayai/status/17136962…
This Could be be such a cool trend! [ use your profile pic +@PurzBeats's pose video + (ipadapter_face + pose controlnet) ]
Ok, i'll try:
1
60
Here was with 3d as a base plate and AI as a render engine back in 2023
nitter.gq/mjayai/status/17307223…
Testing controllable humans expression using a custom blender rig and #SDXL #AnimateDiff for rendering, still alot to refine but preliminary results looks interesting..
@peteromallet
@HelloCivitai
@PurzBeats
#aianimation #aivideo #SDXL #typography #aiartcommunity
28
and its crazy how fast we reached at this point, here is me in 2023 trying to make a video with my likeness, back when i had long hairs:
nitter.gq/mjayai/status/17136962…
This Could be be such a cool trend! [ use your profile pic +@PurzBeats's pose video + (ipadapter_face + pose controlnet) ]
Ok, i'll try:
1
32
Here was with 3d as a base plate and AI as a render engine back in 2023
nitter.gq/mjayai/status/17307223…
Testing controllable humans expression using a custom blender rig and #SDXL #AnimateDiff for rendering, still alot to refine but preliminary results looks interesting..
@peteromallet
@HelloCivitai
@PurzBeats
#aianimation #aivideo #SDXL #typography #aiartcommunity
23
In a world where Creating a High Quality visuals is so cheap, the expensive part will always be storytelling and taste and thats where the Moat is... or I would say, always has been 🔫 ✨
#storybrand #storytelling #seedance #aivideo
Apart from getting Studio Quality outputs, one of the Best things about using Seedance 2.5 is the fact that you can have a consistent concept across different layout sizes!
This means we can truly focus on the fun part! the storytelling!
Ergo, Satisfying the artist and the clients in a single system. what has been one insight you have experienced working with ai commercials, which wasn't possible in a traditional pipeline?
prompt below 👇
#Seedance2_5 #ai #photography #aiartcommunity #commercial #Storytelling
🤖 Made with AI
15-second vertical beauty commercial for MOHMAYA cosmetics. Editorial, minimal, tactile. Two female models (Model A and Model B) in plum and dusty-rose knitwear against a seamless dusty-mauve studio sweep. The whole film is one hue in four values: mauve backdrop, plum and rose knits, frosted mauve packaging with brushed silver caps. Warm skin is the only warm break, and a single berry accent (blush, liner, gloss) is the only saturated colour. Lighting stays the same throughout: one large soft frontal source just above the lens, near shadowless, real skin texture and freckles kept with no smoothing. In macros it rakes slightly so the product reads as raised and glossy. Mostly locked-off camera, eye level. Rhythm: one long still opening, then a fast burst of hard cuts, easing back to stillness, then a graphic endcard.
[0–3s] Chest-up duo portrait, straight on at eye level. Model A and Model B cheek to cheek, Model A slightly forward on the left two-thirds, Model B's face overlapping at the right edge so the two heads read as one shape. Both look straight into the lens with no expression, bare-faced apart from a faint berry note on the lips. Almost no movement: Model A's chin lifts a fraction, Model B's eyes slide toward Model A without her head turning, and Model A's mouth softens slightly without smiling. A very slow push-in, about 8%, that reads as stillness. Small bone-white text "MOHMAYA" in the lower third, then "Colour, worn close.", then it fades out.
[3–4.5s] Hard cut. Product hero: a frosted mauve cream-blush tube with a silver dome cap, held upright in Model A's hand, centred against the empty mauve sweep, hand entering from the lower right. Her thumb rolls the tube a quarter turn so the MOHMAYA wordmark faces the lens, and a specular highlight travels along the barrel. Her other hand enters low with the cap off and a bead of berry cream on the fingertip. Static camera.
[4.5–6.5s] Rapid hard cuts, extreme macro on Model A's freckled cheekbone, lens level with the skin. Three raised dots of berry cream pressed onto the cheek, each casting a tiny shadow. Two fingertips blend them outward into a soft, satiny field. The fingers lift and the colour holds. Tightest shot: the soft graded edge where berry meets bare skin, freckles showing through.
[6.5–8s] Hard cut. Model A in three-quarter close-up, finished cheek to the lens, eyes down. Cut wider to a medium shot: Model B leans in from the right and turns her cheek to the light, comparing. Both carry the same berry flush. Model A's hand enters frame right holding a lip pencil.
[8–9s] Hard cut. Second product hero, framed exactly like the blush shot: a mauve MOHMAYA lip liner with a deep berry tip, held upright, hand low right, empty sweep behind. It rotates to show the wordmark, and a matching gloss tube rises into frame beside it.
[9–11.5s] Hard cuts, extreme macro on the mouth, framed horizontally. The pencil traces the cupid's bow, then slowly closes the line along the lower lip and lifts away, leaving a clean berry outline. A gloss wand lays wet berry into the centre of the lower lip, and the gloss catches a bright specular the matte liner doesn't. The lips press together once and part, and the gloss evens out. Deepest, richest colour of the film.
[11.5–13s] Hard cut. Model B in a clean frontal close-up, finished cheek and lips, eyes to the lens, the first fully finished portrait. A very slow push-in that completes and holds.
[13–14s] Hard cut. Payoff duo portrait matching the opening frame exactly (same placement, distance, lens and light), but now both models are made up and both smile for the first time. Model B lets out one brief, natural laugh.
[14–15s] Hard cut to a flat graphic endcard with zero colour: horizontal black and white bands edge to edge, "MOHMAYA" in widely spaced small caps across the centre band, and a smaller "SHOP NOW" band directly beneath. Static hold to the end.
Avoid: beauty-filter skin smoothing, extra people, handheld shake, lens flares, colour outside the mauve/plum/berry palette, warped or misspelled text, changes to product shape between shots.
123
15-second vertical beauty commercial for MOHMAYA cosmetics. Editorial, minimal, tactile. Two female models (Model A and Model B) in plum and dusty-rose knitwear against a seamless dusty-mauve studio sweep. The whole film is one hue in four values: mauve backdrop, plum and rose knits, frosted mauve packaging with brushed silver caps. Warm skin is the only warm break, and a single berry accent (blush, liner, gloss) is the only saturated colour. Lighting stays the same throughout: one large soft frontal source just above the lens, near shadowless, real skin texture and freckles kept with no smoothing. In macros it rakes slightly so the product reads as raised and glossy. Mostly locked-off camera, eye level. Rhythm: one long still opening, then a fast burst of hard cuts, easing back to stillness, then a graphic endcard.
[0–3s] Chest-up duo portrait, straight on at eye level. Model A and Model B cheek to cheek, Model A slightly forward on the left two-thirds, Model B's face overlapping at the right edge so the two heads read as one shape. Both look straight into the lens with no expression, bare-faced apart from a faint berry note on the lips. Almost no movement: Model A's chin lifts a fraction, Model B's eyes slide toward Model A without her head turning, and Model A's mouth softens slightly without smiling. A very slow push-in, about 8%, that reads as stillness. Small bone-white text "MOHMAYA" in the lower third, then "Colour, worn close.", then it fades out.
[3–4.5s] Hard cut. Product hero: a frosted mauve cream-blush tube with a silver dome cap, held upright in Model A's hand, centred against the empty mauve sweep, hand entering from the lower right. Her thumb rolls the tube a quarter turn so the MOHMAYA wordmark faces the lens, and a specular highlight travels along the barrel. Her other hand enters low with the cap off and a bead of berry cream on the fingertip. Static camera.
[4.5–6.5s] Rapid hard cuts, extreme macro on Model A's freckled cheekbone, lens level with the skin. Three raised dots of berry cream pressed onto the cheek, each casting a tiny shadow. Two fingertips blend them outward into a soft, satiny field. The fingers lift and the colour holds. Tightest shot: the soft graded edge where berry meets bare skin, freckles showing through.
[6.5–8s] Hard cut. Model A in three-quarter close-up, finished cheek to the lens, eyes down. Cut wider to a medium shot: Model B leans in from the right and turns her cheek to the light, comparing. Both carry the same berry flush. Model A's hand enters frame right holding a lip pencil.
[8–9s] Hard cut. Second product hero, framed exactly like the blush shot: a mauve MOHMAYA lip liner with a deep berry tip, held upright, hand low right, empty sweep behind. It rotates to show the wordmark, and a matching gloss tube rises into frame beside it.
[9–11.5s] Hard cuts, extreme macro on the mouth, framed horizontally. The pencil traces the cupid's bow, then slowly closes the line along the lower lip and lifts away, leaving a clean berry outline. A gloss wand lays wet berry into the centre of the lower lip, and the gloss catches a bright specular the matte liner doesn't. The lips press together once and part, and the gloss evens out. Deepest, richest colour of the film.
[11.5–13s] Hard cut. Model B in a clean frontal close-up, finished cheek and lips, eyes to the lens, the first fully finished portrait. A very slow push-in that completes and holds.
[13–14s] Hard cut. Payoff duo portrait matching the opening frame exactly (same placement, distance, lens and light), but now both models are made up and both smile for the first time. Model B lets out one brief, natural laugh.
[14–15s] Hard cut to a flat graphic endcard with zero colour: horizontal black and white bands edge to edge, "MOHMAYA" in widely spaced small caps across the centre band, and a smaller "SHOP NOW" band directly beneath. Static hold to the end.
Avoid: beauty-filter skin smoothing, extra people, handheld shake, lens flares, colour outside the mauve/plum/berry palette, warped or misspelled text, changes to product shape between shots.
1
88
Thanks @WaterrrMiao for the access.
I used it to make a short Halloween explainer on why we actually celebrate on October 31. a funny and informative take on this. 🎃
It’s a Kurtkazart-style piece using @HyperFrames_ . With a few iterations It turned out really great! I honestly didn't expect such quality tbh.
what made the difference is the control path in the hyperframes studio. I could adjust the pacing and the jokes, shape, motion after the first pass, without starting over.
Here’s the result:
A prompt lowers where your codebase rot starts. It doesn't change how fast it rots.
What 7 months of building with AI agents taught me about guardrails that actually refuse, including one that still says no at 90% context.
I built a music tool where you code the track and the visuals react live 🎶
Strudel for the music, a timeline, step grid and piano roll for producing, and Hydra, p5 and GLSL for the visuals.
Built on the shoulders of some amazing open-source tools: Strudel, Hydra and @p5xjs 🙏 #livecoding #algorave
This is Stave 👇
Built on the shoulders of some amazing open-source tools: Strudel, Hydra and @p5xjs 🙏 #livecoding #algorave
17
Mjay retweeted
🔥 Calling all #algorave , #livecoding , and #creativecoding communities!
Give stave.live a spin, push its limits, and share what you create.
Let's see what artists can build with code, sound, and visuals.
🔗 stave.live
Props to @mjayai 🙌
Mjay retweeted
great work! I also tested automated explainer video creation an year back, glad to see its growing fast.
For the research part, I personally wouldn't blindly rely on the agents, I believe its better to proof read and add your own personal experience to the final draft before releasing.
In a tug of war between automation and taste in a creative endevour , Taste almost always wins in the end due to its intent being personal instead of it being the weighted average.