Same prompt on 21 image models. And all of them are for free and you can use them right now.
You can send me your prompt in the comments and I will run it on all of them.
Full-size pictures and the prompt are in the replies ↓
Prompt:
A cinematic film still, a step-printed night look: the still parts sharp while everything in motion smears into stuttering streaks, saturated coloured light, humid air, cramped framing, heavy grain; wistful, forlorn, bittersweet. A strictly limited palette of deep teal, muted teal and soft peach, with warm orange as the only accent; every surface and the light stay within these colours. Medium-long shot, cut at mid-thigh by the bottom edge of the frame, three-quarter view, 600mm super-telephoto lens, extreme compression, the far distance looming enormous behind the small figure, asymmetrical balance, dark mass on one side. Low-key warm tungsten light as a high 45-degree key light, high contrast. A baby goat with floppy ears and tiny hooves, caught mid-hop with every hoof off the ground and its ears flying in the gravel lot beneath an elevated expressway among massive graffiti-covered concrete pillars and abandoned shopping carts. It is in the right third of the frame, facing the open left side, one large dark mass filling the left two-thirds. A freestanding wooden door frame stands in the scene, and through it lies a snowy mountain summit under a hard deep teal sky, loose snow blowing through the gap. Chromatic aberration toward the corners, stroboscopic multiple exposure.
FLUX.3 Klein · seed 7344
Prompt:
A music video still, a lush saturated-glamour look: dense jewel-bright colour on every designed surface, hard bright key light, rich deep blacks, a soft glow blooming on the highlights; defiant, fierce, unbowed. A strictly limited palette of deep plum, carmine red, burnt orange and dark teal, with tangerine orange as the only accent; clothes, surroundings and light stay within these colours. Medium close-up, chest up, head-on frontal shot, 85mm portrait lens, shallow depth of field, negative space, most of the frame one plain empty field. Hard flickering warm orange light, about 1900K from the front, crisp shadows, specular highlights, a few glowing orange embers floating in the air. A breathtakingly beautiful baker in her early thirties, a slim womanly build with a large full bust and deep cleavage, fully clothed in a tight low-cut flour-dusted burnt orange T-shirt and a canvas apron, a tray of hedgehog-shaped bread rolls with almond spikes and raisin eyes held out in both hands in a rural Mexican kitchen with dark teal and burnt orange tiles around an adobe hearth and cut-paper garlands. She fills the frame, the top of her head cut off by the top edge. Two people in their work clothes sit together in a giant porcelain teacup big enough to sit in, steam rising around them, chatting warmly. Hyper-saturated colour.
Qwen-Image 2.1 · seed 311029
Setup: ComfyUI, all local. 896x1120, seed 401101, each model at its default steps. Seconds per picture, model loading included when the model was not loaded yet. Ideogram 4 got the same scene as a JSON caption.
The prompt, word for word:
A real photo of a strikingly beautiful woman in her late twenties sitting at the counter of a tiny night noodle stall in heavy rain, three big orange koi fish swimming through the air around her head. She has sharp cheekbones, dark brows, dark eyes with a graphic white liner wing, a glossy deep-red lip and a platinum-white blunt bob. Her mouth is closed and she looks straight into the camera, deadpan, unimpressed. She wears a glossy black hooded raincoat pushed back off her head, and holds a pair of wooden chopsticks in her right hand over a steaming bowl in front of her on the counter. Her face fills the centre of the frame. Behind her is a dark wet street with a single red paper lantern hanging from the stall roof, steam, and falling rain streaks. The koi are bright orange with white patches, the largest one beside her left ear. Hard warm light from the lantern above-left lights her face and the koi, with deep navy shadows everywhere else. Seen from the front at her eye level, close. Deep navy, orange and white are the only colours, besides black and red.
A woman at a rainy noodle stall, with koi swimming through the air. Same prompt on 21 AI image models, and almost every one got the fish.
All free, and you can use them right now. Comment your prompt and I'll run it on all of them.
The pictures and the prompt are in the reply ↓
@akatz_ai Tried V1.5 with your workflow at your settings, on two clips of my own. Both swaps kept the whole performance, and the faces held to the end. About 9 minutes per 10 s clip. Happy to test the Person Remover the same way if that's useful.
Swap the person in a video for someone from one picture: akatz-ai's Character Swap LoRA for MiniMax H3 kept the performance, room and light; only who's in it changed.
Top: my clips. Bottom: the swaps.
Do the faces match the pictures in the corner?
@speridlabs Tried Idol in ComfyUI, Comfy's template at its defaults. The fox looks great from the front, but the side it never saw comes out smeared, with an extra leg. And my rancher's gold rim light turned everything behind him gold. Am I doing it wrong, or is it a known issue?
ONE picture in, a 3D model out. Idol, a new free image-to-3D model, says it fills in the sides you never showed it.
The catch: my rancher's gold rim light turned the calf and his chaps gold. In even light the calf stayed black.
Does the fox's far side look right to you?
Setup: ComfyUI, all local. 896x1120, seed 301030, each model at its default steps. Seconds per picture, model loading included when the model was not loaded yet. Ideogram 4 got the same scene as a JSON caption.
The prompt, word for word:
A real photo of a strikingly beautiful woman in her early thirties lying back in a tangerine-orange dental chair while a full-grown gorilla in mint-green scrubs and a white face mask leans over her to check her teeth, both completely deadpan. She has sharp cheekbones, dark brows, dark brown eyes with a graphic cobalt-blue liner wing and a glossy deep-red lip; her hair is an electric cobalt-blue bob. She wears a sleeveless black ribbed top under a white paper bib clipped at her collarbones. Her mouth is closed and her eyes look straight at the viewer, cool and unimpressed. Her head rests on the chair's headrest and fills the left-centre of the frame. The gorilla's huge black hands hold a small steel dental mirror in its right hand and a curved steel probe in its left, its dark eyes above the mask turned toward her mouth, its head filling the right third of the frame. Behind them a bright white-mint clinic wall with a plain white cabinet and a mint-green privacy curtain. Hard, bright white light from straight in front throws crisp shadows on the wall. Seen from the foot of the chair, at her eye level, close. Orange, cobalt blue and mint green are the only colours, besides black fur and white.
Is a gorilla a good dentist? Same prompt on 21 AI image models, and every one put him in scrubs.
All free, and you can use them right now. Comment your prompt and I'll run it on all of them.
The pictures and the prompt are in the reply ↓
Prompt:
A digital oil painting, a muted concept-painting look: broad confident brushstrokes, big simple shapes that read at a glance, flat diffuse light, low saturation everywhere, quiet precise everyday detail; lonely, quiet, distant. A strictly limited palette of inky black, gunmetal grey, drained khaki grey and silvery white; clothes, surroundings and light stay within these tones. Full shot, head to toe, the whole figure filling the height of the frame, high-angle shot, 14mm ultra-wide lens up close, wide-angle distortion, silhouette against a bright plain background, one clean dark shape. Flat overcast cool daylight, low contrast, wind driving gray sheets of rain sideways, loose cloth bent hard one way, every surface streaming wet. A strikingly attractive future courier in her late twenties with a thin chrome implant line along one temple, in a cropped padded jacket with drained khaki grey piping and cargo trousers, a sealed hard-shell parcel strapped to her back, bending forward from the hips toward the lens with both hands on the knees, the head and shoulders swollen large near the lens, the legs tapering away below on a long narrow pier of grey wooden planks running out over flat water to a small hut. She is in the right third of the frame, open water filling the left two-thirds.
Z-Image Turbo · seed 211030
Prompt:
A digital oil painting, a muted concept-painting look: broad confident brushstrokes, big simple shapes that read at a glance, flat diffuse light, low saturation everywhere, quiet precise everyday detail; lonely, quiet, distant. A strictly limited palette of inky black, gunmetal grey, drained khaki grey and silvery white; clothes, surroundings and light stay within these tones. Full shot, head to toe, the whole figure filling the height of the frame, high-angle shot, 14mm ultra-wide lens up close, wide-angle distortion, silhouette against a bright plain background, one clean dark shape. Flat overcast cool daylight, low contrast, wind driving gray sheets of rain sideways, loose cloth bent hard one way, every surface streaming wet. A strikingly attractive future courier in her late twenties with a thin chrome implant line along one temple, in a cropped padded jacket with drained khaki grey piping and cargo trousers, a sealed hard-shell parcel strapped to her back, absorbed in a small everyday task, both hands busy on a long narrow pier of grey wooden planks running out over flat water to a small hut. She is in the right third of the frame, open water filling the left two-thirds.
Z-Image Turbo · seed 211030
@mapte_fireball You were right, shorter clips keep it. At 2.3 s, 0 gave exactly 2 jabs and a kick, +4 still added 2 punches. At 1.6 s both gave just the combo, and the kick never got cut off. So +4 only adds moves when there's time left over. Thanks for the idea!
@jun_kongo Tried it, and it works better: the head swap kept the neon light on her face, which the edit model lost. You were right about small faces: the neon one barely changed at my usual size, at 1344x1728 it came much closer to the painting. And in seconds, not minutes. Thanks!
@AnkitKumar65837 Checked it on my files and you're right: the normal load keeps about a quarter of the 4-step LoRA and adds noise to every layer. With Load LoRA (Bypass) the clip changed a lot, but the face still smears without the refine pass. So that's not it, at least on this seed.