Every model, what it is for, and what it costs

OpenAI's newest image model: it follows an instruction to the letter and writes legible text.

Use it for: layouts with copy, product mockups, anything with exact instructions

≈ 85 cr at defaults · from 3 cr

64 cr · 113 sA rain-soaked cyberpunk night market, steam rising from food stalls, holographic signs in Japanese and Russian, a woman in a clear raincoat looking back at the camera, cinematic wide shot.85 cr · 141 sCute 3D character render of a fox astronaut in a tiny orange spacesuit, Pixar style, soft studio lighting, pastel background.

The quickest way to see a prompt at all — rough, instant, and made for trying again.

Use it for: first sketches, checking an idea, hunting for a direction

≈ 2 cr at defaults · from 2 cr

2 cr · 6 sRetro film still: a couple in a 1970s diner booth at night, neon reflections, film grain, Kodak colors.2 cr · 18 sa red vintage bicycle leaning against a white stucco wall, bright midday sun, hard shadows, Mediterranean street

An open model with a loose, painterly hand — forgiving of a vague prompt.

Use it for: concept art, abstract backgrounds, moodboards

≈ 2 cr at defaults · from 2 cr

The workhorse: clean renders with composition you can predict, from text or from a photo.

Use it for: everyday image work, illustration, product shots

≈ 10 cr at defaults · from 10 cr

10 cr · 5 sAerial photo of a Norwegian fjord at dawn, a single red cabin on the shore, mirror-still water, low clouds, muted colors.10 cr · 15 sneon-lit alley after rain, reflections on wet pavement, a lone figure with an umbrella, cyan and magenta signage, cinematic noir, 35mm, film grain

Stable Diffusion at its largest — holds a composition and keeps lettering readable.

Use it for: posters with text, typography, layouts

≈ 26 cr at defaults · from 26 cr

26 cr · 47 sAn astronaut looking out of a space station window at Earth, reflections on the visor, cinematic, volumetric light.26 cr · 250 soil painting of a stormy sea at night, lighthouse beam cutting through rain, thick impasto brushstrokes, dramatic chiaroscuro, museum quality

Google's quick editor: natural light, and edits that leave the rest of the photo alone.

Use it for: photo retouching, putting a product in a scene, quick realism

≈ 16 cr at defaults · from 16 cr

Nano Banana with the controls opened up — shape, scale, and a strip of photos to work from.

Use it for: composites from several photos, brand visuals, print-ready art

≈ 32 cr at defaults · from 32 cr

32 cr · 16 sOverhead food photo: a steaming bowl of tonkotsu ramen with a soft egg, chashu and scallions, dark wooden table, chopsticks, moody light.32 cr · 16 sLoose watercolor painting of a Lisbon street with a yellow tram, warm afternoon light, splashes and white paper showing through.

FLUX Pro, quicker and sharper, with the shape of the frame in your hands.

Use it for: advertising, editorial scenes, detailed renders

≈ 16 cr at defaults · from 16 cr

FLUX at its most photographic: grain, depth and skin that reads as a camera, not a render.

Use it for: photoreal portraits, large prints, campaign visuals

≈ 24 cr at defaults · from 24 cr

24 cr · 11 sArchitectural photo of a sunlit loft interior, concrete ceiling, oak floor, a green velvet sofa, tall windows, dust in the light, wide angle.24 cr · 10 sSports photo: a sprinter exploding out of the starting blocks, frozen motion, dust kicked up, stadium lights, telephoto, dramatic.

Google's top image model — what we reach for when text and faces must survive an edit.

Use it for: text on an image, consistent characters, complex edits

≈ 60 cr at defaults · from 60 cr

60 cr · 22 sStudio fashion portrait of a woman with braided hair and gold hoop earrings, deep red backdrop, dramatic side light, editorial, Hasselblad look.60 cr · 26 sLuxury cosmetics product shot: a frosted glass serum bottle on white marble, soft shadow, single eucalyptus sprig, morning light, minimal.

The previous OpenAI model — still strong on long, specific prompts and on lettering.

Use it for: infographics, mockups, images with captions

≈ 54 cr at defaults · from 4 cr

ByteDance's flagship: motion that stays coherent over a long take, behind a strict content filter.

Use it for: long takes, complex scenes, most video work

≈ 946 cr at defaults · from 353 cr

946 cr · 406 sHandheld tracking shot: a boxer slumps onto the stool in his corner between rounds, sweat flying, his coach leaning in and shouting, ringside lights flaring, crowd blurred behind, gritty 35mm film look.946 cr · 164 sA paper boat drifts down a rain-filled gutter, camera tracks alongside at water level, soft overcast light, gentle ripples, calm and nostalgic.

Seedance's draft tier — text in, a clip out, cheap enough to run again and again.

Use it for: storyboards, previews, cheap iterations

≈ 78 cr at defaults · from 14 cr

78 cr · 36 sA husky puppy bounding through fresh deep snow straight toward the camera, low angle, snow spraying, bright winter sun.78 cr · 31 sTime-lapse of a thunderstorm rolling over a golden wheat field, towering clouds, lightning strikes on the horizon, wind bending the wheat.

The classic Kling — steady motion and physics that behave, from text or a photo.

Use it for: simple motion, a product turning, a first animation of a photo

≈ 112 cr at defaults · from 112 cr

112 cr · 208 sProduct shot: a white and orange running sneaker rotating slowly on a black surface, rim light, water droplets on the mesh, studio.

Seedance's first generation at full quality — a polished clip straight from text.

Use it for: polished short clips, advertising, social video

≈ 243 cr at defaults · from 20 cr

Kling 3 with sound of its own, a closing frame, and no objection to real faces.

Use it for: clips with people, dialogue scenes, animating a portrait

≈ 252 cr at defaults · from 101 cr

252 cr · 59 sDrone shot rising over a winding coastal cliff road at sunrise, a red vintage convertible driving along it, waves crashing below, golden light, smooth and cinematic.252 cr · 80 sClose-up portrait of an elderly fisherman laughing on a boat deck, sea spray, golden hour backlight, shallow depth of field, gentle handheld camera.

Kling 3's senior tier — finer detail and a camera that wanders less.

Use it for: hero shots, advertising, anything that must look finished

≈ 336 cr at defaults · from 135 cr

336 cr · 287 sExtreme macro: espresso pouring from a portafilter into a clear glass cup, crema swirling in slow motion, warm studio light, black background.

The Kling that keeps a character recognisable from shot to shot, sound included.

Use it for: characters across shots, scenes built from references, clips with sound

≈ 168 cr at defaults · from 101 cr

Kling o3's senior tier — steadier motion and finer detail in every mode it has.

Use it for: hero shots with recurring characters, advertising, finished pieces

≈ 224 cr at defaults · from 135 cr

The previous Seedance flagship — cinematic motion with sound, photos addressed as @Image.

Use it for: cinematic scenes, compositions from several photos, high-resolution deliverables

≈ 605 cr at defaults · from 218 cr

605 cr · 182 sA dancer in a yellow raincoat moving through a rain-soaked Tokyo alley at night, neon signs reflecting in puddles, slow motion, water droplets catching the light, cinematic 35mm, vertical frame.

Seedance's quick branch — the same read of a prompt, a lighter picture, made for drafting.

Use it for: drafts of Seedance scenes, volume work, checking a shot before the full take

≈ 484 cr at defaults · from 175 cr

Google Veo films rather than animates, and it speaks: dialogue, ambience, a sense of lens.

Use it for: dialogue scenes, realistic footage, first-to-last-frame shots

≈ 1280 cr at defaults · from 320 cr

Google's fast video model — it also edits a clip you already have.

Use it for: quick clips, edits of existing video, scenes from several photos

≈ 320 cr at defaults · from 36 cr

200 cr · 36 sA skateboarder lands a kickflip in an empty concrete parking garage, sparks from the wheels, slow motion, harsh fluorescent light, low angle.

Veo's cheap tier — the same sound and camera sense, starting from your photo.

Use it for: drafts, social clips, trying an idea before the full Veo

≈ 160 cr at defaults · from 48 cr

Alibaba Wan — long, patient takes and a strip of photos to build a scene from.

Use it for: longer scenes, sequences from several references, unhurried shots

≈ 400 cr at defaults · from 40 cr

Wan's senior tier — the same modes with cleaner motion and more detail held.

Use it for: finished pieces, work where the extra sharpness shows, client deliverables

≈ 560 cr at defaults · from 55 cr

xAI Grok Imagine — quick and literal; it rewards a short, concrete prompt.

Use it for: quick ideas, a single clear action, social posts

≈ 336 cr at defaults · from 32 cr

MiniMax H3 holds a face steady through a clip, and renders at high resolution.

Use it for: recurring characters, clips that need resolution, faces in motion

≈ 260 cr at defaults · from 100 cr

The larger MiniMax H3 — steadier motion and a firmer grip on a character.

Use it for: clips that must hold together, recurring characters, longer actions

≈ 160 cr at defaults · from 100 cr

The quickest way to see an idea move at all — rough, and made for trying again.

Use it for: first drafts, checking a prompt, throwaway attempts

≈ 80 cr at defaults · from 50 cr

Veo's quick branch — the same eye, a softer picture, made for drafting a scene.

Use it for: drafts of Veo scenes, volume work, checking a shot before the full take

≈ 480 cr at defaults · from 160 cr

360 cr · 78 sA street musician playing saxophone under a stone bridge at night, passing car headlights sweeping across wet cobblestones, soft saxophone melody and city ambience.

ElevenLabs — what we reach for when the reading must not sound synthetic.

Use it for: narration, advertising, character voices

40 cr per 1,000 characters

MiniMax — strong Russian, with presets for emotion and a hand on the pace.

Use it for: voice-over in Russian, expressive reads, social clips

40 cr per 1,000 characters