Every model, and
what it actually costs
37 generation endpoints, 15 billed per image and 22 per second. Every price below is read out of the same code that runs the generations, at the provider's list rate. A sale only ever lowers it: when the provider discounts a model, the app prices your run at the sale rate before you start.
Image edit
Image to image. You supply one or more reference images and a prompt, and the model returns a new image of the same subject. This is the lane every persona workflow runs on, because the reference is what holds the face steady.
| Model | Provider | From, per image | Tiers |
|---|---|---|---|
| OpenAI GPT Image 2 | OpenAI | $0.02 | low, 1k $0.02low, 2k $0.03low, 4k $0.04medium, 1k $0.07+5 more |
| ByteDance Seedream v4 | ByteDance | $0.027 | flat $0.027 |
| Google Nano-Banana | $0.038 | flat $0.038 | |
| ByteDance Seedream v4.5 | ByteDance | $0.04 | flat $0.04 |
| ByteDance Seedream v5.0 Pro | ByteDance | $0.045 | 1k $0.0452k $0.09 |
| Google Nano-Banana 2 Fast | $0.045 | 2k $0.0454k $0.05 | |
| Google Nano-Banana 2 | $0.045 | 0.5k $0.0451k $0.072k $0.1054k $0.14 | |
| Flux 2 Turbo | WaveSpeed | $0.048 | flat $0.048 |
| Google Nano-Banana Pro | $0.14 | 1k $0.142k $0.144k $0.24 | |
| Google Nano-Banana Pro Ultra | $0.15 | 4k $0.158k $0.18 |
Text-to-image
Prompt in, image out, with no reference image at all. Cheaper per image than the edit endpoints in most cases, and the right lane for backgrounds, products and anything where nothing has to stay consistent between runs.
| Model | Provider | From, per image | Tiers |
|---|---|---|---|
| OpenAI GPT Image 2 | OpenAI | $0.01 | low, 1k $0.01low, 2k $0.02low, 4k $0.03medium, 1k $0.06+5 more |
| Flux 2 Turbo | WaveSpeed | $0.04 | flat $0.04 |
| ByteDance Seedream v5.0 Pro | ByteDance | $0.045 | 1k $0.0451.5k $0.0452k $0.09 |
| Google Nano-Banana 2 | $0.045 | 0.5k $0.0451k $0.072k $0.1054k $0.14 | |
| Google Nano-Banana Pro | $0.14 | 1k $0.142k $0.144k $0.24 |
Image-to-video
One still image becomes a clip. Billed per output second, so duration is the dominant term in the bill and the resolution tier is usually the second. Capabilities vary far more than the prices do.
| Model | Provider | From, per second | Tiers |
|---|---|---|---|
| Veo 3.1 Lite | $0.05 | 720p $0.051080p $0.08 | |
| Grok Imagine Video | xAI | $0.05 | flat $0.05 |
| Seedance 2.0 Mini | ByteDance | $0.06 | 480p $0.06720p $0.121080p $0.304k $0.60 |
| LTX-2 Pro | Lightricks | $0.06 | flat $0.06 |
| Kling 3.0 Std | Kuaishou | $0.084 | flat $0.084 |
| Veo 3.1 Fast | $0.10 | 720p $0.101080p $0.104k $0.25720p with audio $0.15+2 more | |
| WAN 2.7 | Alibaba | $0.10 | 720p $0.101080p $0.15 |
| Sora 2 | OpenAI | $0.10 | flat $0.10 |
| Kling 3.0 Pro | Kuaishou | $0.112 | flat $0.112 |
| Seedance 2.0 | ByteDance | $0.12 | 480p $0.12720p $0.241080p $0.604k $1.20 |
| Seedance 2.5 | ByteDance | $0.18 | 480p $0.18720p $0.361080p $0.904k $1.80 |
| Seedance 2.5 Turbo | ByteDance | $0.20 | 720p $0.201080p $0.22 |
Motion control
A source clip supplies the movement and a still supplies the subject. Billed against the SOURCE clip length rather than a duration you choose, with a floor and a cap that make short and long clips behave unlike every other lane here.
| Model | Provider | From, per second | Tiers |
|---|---|---|---|
| Kling 2.6 Std | Kuaishou | $0.07 | flat $0.07 |
| Kling 2.6 Pro | Kuaishou | $0.112 | flat $0.112 |
| Kling 3.0 Std | Kuaishou | $0.126 | flat $0.126 |
| Kling 3.0 Pro | Kuaishou | $0.168 | flat $0.168 |
Talking avatar
A photo plus an audio track becomes a lip-synced clip. Duration is not a control: it tracks the audio you upload. That makes the per-model minimum billed unit and maximum clip length the two numbers that decide what you actually pay.
| Model | Provider | From, per second | Tiers |
|---|---|---|---|
| InfiniteTalk Fast | WaveSpeed | $0.015 | flat $0.015 |
| InfiniteTalk | WaveSpeed | $0.03 | 480p $0.03720p $0.06 |
| SkyReels V3 | WaveSpeed | $0.03 | 480p $0.03720p $0.06 |
| LongCat Avatar 1.5 | WaveSpeed | $0.04 | 480p $0.04720p $0.08 |
| Kling V2 Avatar Std | Kuaishou | $0.056 | flat $0.056 |
| OmniHuman 1.5 | ByteDance | $0.16 | flat $0.16 |
Where these numbers come from
Each price is transcribed from the model's own live pricing formula on the provider's page, not from its documentation and not from the listing grid. The listing shows the sale rate whenever a model is discounted, and sales end without notice, so the list rate is what appears here. The app itself shows the sale rate while one runs.
The figures are list price for the compute alone. Personifex does not resell it. You bring your own provider key, the provider bills you directly, at these rates or below them during a sale, and the subscription is a flat fee for the workflow around it.
Caveat Providers change prices without notice. Every figure here is compared with the provider's own pricing records each week, and a change is corrected in the app and on this page together.