01Realistic portrait
Gemini 3 Pro Image
Nano Banana Pro · Google
The steadiest portrait model. Skin tone, eye highlights, fabric folds hold up under zoom. The more you write about light, the closer it lands.
PhotorealPortraitsEditorial
ISSUE 002 · WINTER · 2026
Image&
Video.
Write one line — it can be a photo-realistic image, or a cinematic shot. The frontier models live on the same canvas, swap any time.
— Type to start —
Model Lineup / The Lineup
Onboarded this issue · Up to date
Image series · IMAGE13 models
- Gemini 3 Pro ImageNano Banana Pro
- Gemini 3.1 FlashNano Banana 2
- Gemini 2.5 FlashNano Banana
- Imagen 4.0Google
- GPT-Image-2OpenAI
- GPT-Image 1 / 1.5OpenAI
- Seedream 5.0 LiteByteDance
- Seedream 4.5ByteDance
- Seedream 4.0ByteDance
- Qwen Image 2.0Alibaba
- Z-Image-TurboZhipu
- Grok Imagine 1.0xAI
- Wan 2.7 ImageAlibaba
Video series · VIDEO19 models
- Sora 2OpenAI
- VEO 3Google
- Doubao Seedance 2.0ByteDance
- Doubao Seedance 1.5 ProByteDance
- Doubao Seedance 1.0 ProByteDance
- MiniMax Hailuo 2.3海螺
- MiniMax Hailuo 02海螺
- Kling v3 Omni可灵
- Kling v3可灵
- Kling 2.6可灵
- Kling Video O1可灵
- Vidu Q3 (pro/turbo)生数
- Vidu Q3 (mix/std)生数
- SkyReels V4SkyReels
- HappyHorse 1.0快马
- Wan 2.7Alibaba
- Wan 2.6 / 2.5Alibaba
- Grok Imagine 1.0xAI
- Omni-Flash-Ext扩展通用
Two big things · The two things
Two worlds on the same canvas.
Image freezes a moment; video lets it move. Different models, but the same language for telling them what to do.
IMAGE
From prompt to image.
Composition, light, in-frame text — write them in natural language. Nano Banana shines at realism, Imagen at editorial poise, Seedream at Chinese in-image text.
Nano Banana ProImagen 4.0GPT-Image-2Seedream 5.0Qwen 2.0
VIDEO
From prompt to motion.
Generate a short clip, or turn a still into a moving shot. Camera, motion, rhythm all live in the prompt — Sora / VEO / Kling each have a voice.
Sora 2VEO 3Kling v3 OmniHailuo 2.3Doubao Seedance 2.0
Editor's Picks
Models have temperaments.
Knowing which to pick is half the time saved.
Same prompt, different model, completely different result. Here are our default picks for common jobs.
02Editorial illustration
Imagen 4.0
Clean palette, tidy composition. For covers, posters, brand assets — start here to find direction, almost no wasted drafts.
PosterIllustrationTidy comp
03Chinese in-frame text
Seedream 5.0
ByteDance
Chinese signs, labels, headlines inside the image — go here first. Vertical, traditional, hand-style all work.
In-frame textChineseProduct
04Cinematic video
Sora 2
OpenAI
Strongest at camera moves, pacing, character consistency. Pick this for narrative shorts — it remembers who the subject is.
CinematicSubject consistencyNarrative
05Realistic video
VEO 3
Best physics. Water, fabric, smoke, glass — anything that has to 'look real' should run here.
PhysicsMaterialsCommercial
06Image-to-video
Kling v3 Omni
可灵 · Kuaishou
Most reliable when turning a still into a clip. First-to-last frame transitions are smooth; also handles start/end frame interpolation.
Image-to-videoKeyframesMotion
Not sure which? The Lab auto-suggests by prompt — you can override any time.
Rate Card
Top up as needed.
Points never expire.
Points are the universal fuel in the Lab. Different models / durations consume differently — the live rate shows in the canvas corner at run time.
TierPointsBonusPrice
Points never expireFailed runs aren't chargedAlipay / CardCommercial use for paying users
FAQ
You'd probably ask.
How are points charged?
Different models and different video durations cost different points. The exact unit price is shown in the canvas corner when you run. Failed runs aren't charged.
Which models? Will they go stale?
We keep adding new ones. Today: image — Gemini Nano Banana family, Imagen 4, GPT-Image, Seedream, Qwen, Z-Image, Grok and more; video — Sora 2, VEO 3, Kling, Hailuo, Doubao Seedance, Vidu, Wan and more.
Who owns the images and videos?
Every run goes to your private library by default; only you can see it. You can download, remix, and publish to the public gallery any time — you control the pace, and can unlist any time.
Can I turn an image into a video?
Yes. Any generated or uploaded image can be used as a first/reference frame for a video model — right-click 'to video' in your private library to switch.
Can I use the outputs commercially?
Paying users hold full usage rights to their generated images and videos, including commercial use. Full terms in the ToS.
Image & Video — one canvas is enough.
30-second signup, free credits land instantly — pick a model, write one line, see what it can make.
Start generating