Clear CortexEvery model. One login, one pot of credit.
7 image models and 8 video models in the same studios, on the same credit — no seat to buy, and you pay for what you actually make. Here is what each one is genuinely better at.
Nano Banana, GPT Image 2, Seedream and Flux.2. What separates them is text, realism, references and resolution.
Nano Banana 2
Google's Gemini 3.1 Flash Image, wired into a canvas.
What it's forNano Banana Pro
Google's flagship image model, on a pay-per-image account.
What it's forGPT Image 2
OpenAI's image model, with the text rendering the others miss.
What it's forSeedream 4.5
ByteDance's 4K model — and it costs the same at every size.
What it's forFlux.2 Klein
The cheap, fast one — for the ninety drafts before the good one.
What it's forFlux.2 Pro
Sharp, photographic detail from Black Forest Labs.
What it's forFlux.2 Max
Black Forest Labs' highest fidelity, per image.
What it's forEvery image model on Clear Cortex
One account, one pot of credit, all 7. Switch model between generations without losing your prompt.
| Model | Maker | Max res | Ratios | Refs | Text in image |
|---|---|---|---|---|---|
| Nano Banana 2 | 4K | 14 | 14 | ||
| Nano Banana Pro | 4K | 10 | 14 | ||
| GPT Image 2 | OpenAI | — | 9 | 16 | |
| Seedream 4.5 | ByteDance | 4K | 18 | 14 | |
| Flux.2 Klein | Black Forest Labs | — | 9 | 4 | |
| Flux.2 Pro | Black Forest Labs | — | 9 | 8 | |
| Flux.2 Max | Black Forest Labs | — | 9 | 8 |
A dash under “Max res” means the model has no resolution setting to choose — it returns its own native size — not that it is limited to a small one. “Text in image” marks the model with per-tier quality control and a background setting — GPT Image 2 — which is the one built for legible type and clean cut-outs.
Seedance 2.5 and 2.0, Kling 3.0, Veo 3.1 and Hailuo 3. Every one takes a first and last frame and can generate its own audio; what separates them is length, resolution and price.
Seedance 2.0 Mini
ByteDance's draft tier for video — clips up to 15 seconds.
What it's forSeedance 2.0
480p to 4K, and up to fifteen seconds of it.
What it's forSeedance 2.5
Up to thirty seconds in a single take.
What it's forKling 3.0 Standard
Kuaishou's everyday tier — three seconds to fifteen, one flat rate.
What it's forKling 3.0 Pro
Every option Standard has, at a third more per second.
What it's forVeo 3.1 Fast
Google's Veo at the fast tier — 4, 6 or 8 seconds, up to 4K.
What it's forVeo 3.1
Google's top Veo tier, on a pay-per-clip account.
What it's forHailuo 3
MiniMax's model — every clip at 2K, five seconds to fifteen.
What it's forEvery video model on Clear Cortex
One account, one pot of credit, all 8. Switch model between clips without losing your prompt.
| Model | Maker | Max res | Clip length | Ratios | Refs | Billed |
|---|---|---|---|---|---|---|
| Seedance 2.0 Mini | ByteDance | 720p | 4–15s | 7 | 4 | per video token |
| Seedance 2.0 | ByteDance | 4K | 4–15s | 7 | 4 | per video token |
| Seedance 2.5 | ByteDance | 720p | 4–30s | 6 | 4 | per video token |
| Kling 3.0 Standard | Kuaishou | 720p | 3–15s | 3 | 0 | per second of video |
| Kling 3.0 Pro | Kuaishou | 720p | 3–15s | 3 | 0 | per second of video |
| Veo 3.1 Fast | 4K | 4–8s | 2 | 0 | per second of video | |
| Veo 3.1 | 4K | 4–8s | 2 | 0 | per second of video | |
| Hailuo 3 | MiniMax | 2K | 5–15s | 6 | 5 | per second of video |
All seven take a first frame and a last frame, and all seven can generate audio with the clip, so neither is a column. “Clip length” is the span the model accepts — Veo takes only 4, 6 or 8 seconds inside its range, the rest take every whole second. “Billed” is the unit the provider charges in: a per-second model's cost is settled before the job runs, a per-token model's moves with the resolution you pick as well as the length.