# Pricing

Source: https://docs.prodia.com/reference/pricing/

Every job is billed as a single product, chosen by the job type and a small number of config parameters (output resolution, duration, number of input images). Prompt, seed, steps and other settings do not change the price.

> tip:
>
> Use [`?price=true`](https://docs.prodia.com/reference/price/) to read back the exact product and dollar amount charged for any job.

## Models

### Video models

- [Kling](https://docs.prodia.com/reference/pricing/#kling)
- [LTX 2.5 Fast](https://docs.prodia.com/reference/pricing/#ltx-25-fast)
- [MiniMax H3 Fast](https://docs.prodia.com/reference/pricing/#minimax-h3-fast)
- [Pruna P-Video](https://docs.prodia.com/reference/pricing/#pruna-p-video)
- [Seedance](https://docs.prodia.com/reference/pricing/#seedance)
- [Sora 2](https://docs.prodia.com/reference/pricing/#sora-2)
- [Veo 3](https://docs.prodia.com/reference/pricing/#veo-3)
- [Wan 2.7](https://docs.prodia.com/reference/pricing/#wan-27)

### Image models

- [FLUX.1](https://docs.prodia.com/reference/pricing/#flux1)
- [FLUX.1 Kontext](https://docs.prodia.com/reference/pricing/#flux1-kontext)
- [FLUX.2 Klein 4B](https://docs.prodia.com/reference/pricing/#flux2-klein-4b)
- [Nano Banana](https://docs.prodia.com/reference/pricing/#nano-banana)
- [Qwen Image Edit Plus](https://docs.prodia.com/reference/pricing/#qwen-image-edit-plus)
- [Recraft](https://docs.prodia.com/reference/pricing/#recraft)
- [SDXL](https://docs.prodia.com/reference/pricing/#sdxl)
- [Seedream](https://docs.prodia.com/reference/pricing/#seedream)

### Utilities

- [Image Utilities](https://docs.prodia.com/reference/pricing/#image-utilities)

## Kling

Applies to the [Kling](https://docs.prodia.com/models/kling/) job types `inference.kling.txt2vid.v1` and `inference.kling.img2vid.v1`.

Priced per video by the `model` version, `mode` (`std` or `pro`, where the model offers both) and `duration` (`"5"` or `"10"` seconds). Aspect ratio and prompt do not change the price, and image-to-video bills the same as text-to-video. `model` defaults to `kling-v1` and `mode` to `std` when omitted.

### Price per video

| Model               | Mode     | Resolution | 5 seconds | 10 seconds |
| ------------------- | -------- | ---------- | --------: | ---------: |
| `kling-v1`          | Standard | 720P       |     $0.15 |      $0.30 |
| `kling-v1`          | Pro      | 1080P      |     $0.55 |      $1.10 |
| `kling-v1-6`        | Standard | 720P       |     $0.25 |      $0.50 |
| `kling-v1-6`        | Pro      | 1080P      |     $0.55 |      $1.10 |
| `kling-v2-master`   | —        | 1080P      |     $1.40 |      $2.80 |
| `kling-v2-1`        | Standard | 720P       |     $0.28 |      $0.56 |
| `kling-v2-1-master` | —        | 1080P      |     $1.40 |      $2.80 |

`kling-v2-master`, `kling-v2-1` and `kling-v2-1-master` bill the same regardless of `mode`.

### Product codes

The `product` field follows the pattern `inference-kling-{v1,v1-6}-{std,pro}-duration-N`, `inference-kling-v2-1-standard-duration-N`, or `inference-kling-{v2-master,v2-1-master}-duration-N`, where `N` is 5 or 10. For example a 10-second `kling-v1-6` video in `pro` mode bills as `inference-kling-v1-6-pro-duration-10`.

## LTX 2.5 Fast

Applies to the LTX 2.5 `fast` tier job types:

- `inference.ltx-2.5.fast.txt2vid.v0`
- `inference.ltx-2.5.fast.img2vid.v0`

Priced per clip by output duration. Fractional durations round up to the next whole second. Output resolution, aspect ratio and seed do not change the price, and image-to-video bills the same as text-to-video.

### Price per clip

| Duration |   Price |
| -------- | ------: |
| 1 s      | $0.0152 |
| 2 s      | $0.0181 |
| 3 s      | $0.0231 |
| 4 s      | $0.0316 |
| 5 s      | $0.0346 |
| 6 s      | $0.0391 |
| 7 s      | $0.0445 |
| 8 s      | $0.0506 |
| 9 s      | $0.0581 |
| 10 s     | $0.0655 |

### Product codes

The `product` field follows the pattern `inference-ltx-2-5-duration-N`, where `N` is the billed duration in seconds (1 to 10). For example a 6-second clip bills as `inference-ltx-2-5-duration-6`.

## MiniMax H3 Fast

Applies to the MiniMax H3 `fast` tier job types:

- `inference.minimax.h3.fast.txt2vid.v1`
- `inference.minimax.h3.fast.img2vid.v1`
- `inference.minimax.h3.fast.ref2vid.v1`

Priced at **$0.08 per second** of output video. Durations run from 4 to 15 seconds and are billed in whole seconds; a job with no `duration` set bills at the 6-second default.

### Price per video

| Duration | Price |
| -------- | ----: |
| 4 s      | $0.32 |
| 5 s      | $0.40 |
| 6 s      | $0.48 |
| 7 s      | $0.56 |
| 8 s      | $0.64 |
| 9 s      | $0.72 |
| 10 s     | $0.80 |
| 11 s     | $0.88 |
| 12 s     | $0.96 |
| 13 s     | $1.04 |
| 14 s     | $1.12 |
| 15 s     | $1.20 |

### Input image surcharge

Input images are counted across the reference images on `ref2vid` jobs and any first or last frame on `img2vid` jobs. Up to 5 input images are included in the per-second price. From the sixth image onwards a flat surcharge is added to the whole video, regardless of duration:

| Input images | Surcharge per video |
| ------------ | ------------------: |
| 0 to 5       |                none |
| 6            |               $0.04 |
| 7            |               $0.08 |
| 8            |               $0.12 |
| 9            |               $0.16 |

For example, a 6-second `ref2vid` job with 7 reference images bills $0.48 + $0.08 = $0.56.

Reference videos on `ref2vid` jobs are billed at the same per-second rate as the output: the input video's duration is added to the output duration before pricing.

### Product codes

The `product` field follows the pattern `inference-minimax-h3-fast[-M-in]-duration-N`, where `M` is the number of input images (omitted when there are none) and `N` is the billed duration in seconds. For example a 6-second job with 7 reference images bills as `inference-minimax-h3-fast-7-in-duration-6`.

## Pruna P-Video

Applies to the P-Video job types:

- `inference.pruna.p-video.txt2vid.v1`
- `inference.pruna.p-video.img2vid.v1`
- `inference.pruna.p-video.aud2vid.v1`

Priced per video by mode (`draft: true` for Draft, otherwise Standard), `resolution` and duration. Aspect ratio and prompt do not change the price. Text- and image-to-video take a `duration` of 5 or 10 seconds; audio-to-video follows the length of the input audio and bills at the 5, 10 or 15 second tier, rounded up.

### Price per video

| Mode     | Resolution | 5 seconds | 10 seconds | 15 seconds |
| -------- | ---------- | --------: | ---------: | ---------: |
| Draft    | 720p       |    $0.025 |     $0.050 |     $0.075 |
| Draft    | 1080p      |    $0.050 |     $0.100 |     $0.150 |
| Standard | 720p       |    $0.100 |     $0.200 |     $0.300 |
| Standard | 1080p      |    $0.200 |     $0.400 |     $0.600 |

### Product codes

The `product` field follows the pattern `inference-pruna-p-video[-1080p][-draft]-duration-N`, where `N` is 5, 10 or 15. 720p and Standard are the defaults and carry no suffix. For example a 10-second Draft video at 1080p bills as `inference-pruna-p-video-1080p-draft-duration-10`.

## Seedance

Applies to the [Seedance](https://docs.prodia.com/models/seedance/) 1.0 job types, in three tiers:

- **Lite** – `inference.seedance.lite.txt2vid.v1` and `inference.seedance.lite.img2vid.v1`
- **Pro** – `inference.seedance.pro.txt2vid.v1` and `inference.seedance.pro.img2vid.v1`
- **Pro Turbo** – `inference.seedance.proturbo.txt2vid.v1` and `inference.seedance.proturbo.img2vid.v1`

Priced per video by `resolution` and `duration` (5 or 10 seconds). Aspect ratio, prompt and seed do not change the price, and image-to-video bills the same as text-to-video.

### Price per video

| Tier      | Resolution | 5 seconds | 10 seconds |
| --------- | ---------- | --------: | ---------: |
| Lite      | 480p       |     $0.09 |      $0.18 |
| Lite      | 720p       |     $0.20 |      $0.40 |
| Lite      | 1080p      |     $0.44 |      $0.88 |
| Pro       | 480p       |     $0.12 |      $0.24 |
| Pro       | 1080p      |     $0.61 |      $1.22 |
| Pro Turbo | 480p       |    $0.048 |     $0.096 |
| Pro Turbo | 1080p      |    $0.244 |     $0.488 |

Pro and Pro Turbo do not offer 720p. When omitted, `resolution` defaults to `1080p` and `duration` to 5 seconds.

### Product codes

The `product` field follows the pattern `inference-seedance-{lite,pro,proturbo}-{480p,720p,1080p}-duration-{5,10}`. For example a 10-second Pro Turbo video at 480p bills as `inference-seedance-proturbo-480p-duration-10`.

## Sora 2

Applies to the [Sora 2](https://docs.prodia.com/models/sora-2/) job types:

- **Sora 2** – `inference.sora-2.txt2vid.v1` and `inference.sora-2.img2vid.v1`
- **Sora 2 Pro** – `inference.sora-2.pro.txt2vid.v1` and `inference.sora-2.pro.img2vid.v1`

Priced per second of output video. `duration` is 4, 8 or 12 seconds (default 4). Sora 2 renders at 720p; Sora 2 Pro takes a `resolution` of `720p` (default) or `1080p`. Aspect ratio, prompt and seed do not change the price, and image-to-video bills the same as text-to-video.

### Price per video

| Model      | Resolution | Per second | 4 seconds | 8 seconds | 12 seconds |
| ---------- | ---------- | ---------: | --------: | --------: | ---------: |
| Sora 2     | 720p       |      $0.10 |     $0.40 |     $0.80 |      $1.20 |
| Sora 2 Pro | 720p       |      $0.30 |     $1.20 |     $2.40 |      $3.60 |
| Sora 2 Pro | 1080p      |      $0.50 |     $2.00 |     $4.00 |      $6.00 |

### Product codes

The `product` field follows the pattern `inference-sora-2-duration-N` for Sora 2 and `inference-sora-2-pro-{720p,1080p}-duration-N` for Sora 2 Pro, where `N` is 4, 8 or 12. For example an 8-second Sora 2 Pro video at 1080p bills as `inference-sora-2-pro-1080p-duration-8`.

## Veo 3

Applies to the [Veo](https://docs.prodia.com/models/veo/) job types:

- **Veo 3 Fast** – `inference.veo.fast.txt2vid.v2` and `inference.veo.fast.img2vid.v2` (and their `v1` equivalents)
- **Veo 3** – `inference.veo.txt2vid.v2` and `inference.veo.img2vid.v2` (and their `v1` equivalents)

Every Veo 3 video is 8 seconds long. Priced per video by tier and whether audio is generated (`generate_audio: true`). Output resolution (720p or 1080p), prompt and seed do not change the price, and image-to-video bills the same as text-to-video.

### Price per video

| Model      | Length    | Without audio | With audio |
| ---------- | --------- | ------------: | ---------: |
| Veo 3 Fast | 8 seconds |         $0.80 |      $1.20 |
| Veo 3      | 8 seconds |         $1.60 |      $3.20 |

### Product codes

The `product` field is `inference-veo` or `inference-veo-fast`, with `-audio` appended when audio is generated. Jobs that set `aspect_ratio` explicitly carry it in the product as well, for example `inference-veo-fast-16_9-audio`.

## Wan 2.7

Applies to the [Wan 2.7](https://docs.prodia.com/models/wan-2-7/) job types:

- **Video** – `inference.wan2-7.txt2vid.v1`, `inference.wan2-7.img2vid.v1` and `inference.wan2-7.vid2vid.v1`
- **Image** – `inference.wan2-7.txt2img.v1` and `inference.wan2-7.img2img.v1`

Video is priced per second of output by `resolution` (`720P`, the default, or `1080P`). `duration` runs from 2 to 15 seconds and is billed in whole seconds, rounded up. Images are a flat price at either `1K` or `2K`. Aspect ratio, prompt, seed and reference inputs do not change the price.

### Video price per second

| Resolution | Per second | 5 seconds | 10 seconds | 15 seconds |
| ---------- | ---------: | --------: | ---------: | ---------: |
| 720P       |     $0.084 |     $0.42 |      $0.84 |      $1.26 |
| 1080P      |     $0.140 |     $0.70 |      $1.40 |      $2.10 |

### Image price

| Output   | Example sizes        |  Price |
| -------- | -------------------- | -----: |
| 1K or 2K | 1024×1024, 2048×2048 | $0.028 |

### Product codes

Video bills as `inference-wan2-7-{720p,1080p}-duration-N`, where `N` is the billed duration in seconds (2 to 15). For example a 5-second 1080P clip bills as `inference-wan2-7-1080p-duration-5`. Images bill as `inference-wan2-7-image`.

## FLUX.1

Applies to the FLUX.1 job types:

- **FLUX.1 \[dev]** – `inference.flux.dev.txt2img.v2`, `inference.flux.dev.img2img.v2`, `inference.flux.dev.inpainting.v2` (and their `v1` equivalents), plus `inference.flux-fast.dev.txt2img.v1`
- **FLUX1.1 \[pro]** – `inference.flux.pro11.txt2img.v1`
- **FLUX1.1 \[pro] Ultra** – `inference.flux.pro11ultra.txt2img.v1` and `inference.flux.pro11ultra.img2img.v1`

FLUX.1 \[dev] is priced per image by step count; the FLUX1.1 \[pro] models are a flat price per image. Output resolution, prompt and seed do not change the price. Volume pricing applies from 1M generations.

### Price per image

| Model                | Steps    |   Price | Volume discount |
| -------------------- | -------- | ------: | :-------------: |
| FLUX.1 \[dev]        | Up to 28 | $0.0200 |       25%       |
| FLUX.1 \[dev]        | 29 to 50 | $0.0240 |       25%       |
| FLUX1.1 \[pro]       | —        | $0.0400 |       25%       |
| FLUX1.1 \[pro] Ultra | —        | $0.0600 |       25%       |

### Product codes

| Model                | Product                                                      |
| -------------------- | ------------------------------------------------------------ |
| FLUX.1 \[dev]        | `inference-flux-dev-steps-28`, `inference-flux-dev-steps-50` |
| FLUX1.1 \[pro]       | `inference-flux-pro11`                                       |
| FLUX1.1 \[pro] Ultra | `inference-flux-pro11-ultra`                                 |

## FLUX.1 Kontext

Applies to the [FLUX.1 Kontext](https://docs.prodia.com/models/flux-kontext/) job types:

- **FLUX.1 Kontext \[dev]** – `inference.flux-fast.dev.kontext.img2img.v1`
- **FLUX.1 Kontext \[pro]** – `inference.flux-kontext.pro.txt2img.v2` and `inference.flux-kontext.pro.img2img.v2` (and their `v1` equivalents)
- **FLUX.1 Kontext \[max]** – `inference.flux-kontext.max.txt2img.v2` and `inference.flux-kontext.max.img2img.v2` (and their `v1` equivalents)

Flat price per image. Output resolution, aspect ratio, prompt and seed do not change the price, and image-to-image bills the same as text-to-image. Volume pricing applies from 1M generations.

### Price per image

| Model                 | Example sizes                 |  Price | Volume discount |
| --------------------- | ----------------------------- | -----: | :-------------: |
| FLUX.1 Kontext \[dev] | 1024×1024, 1568×672, 672×1568 | $0.025 |       25%       |
| FLUX.1 Kontext \[pro] | 1024×1024, 1568×672, 672×1568 | $0.040 |       25%       |
| FLUX.1 Kontext \[max] | 512×512, 512×768, 640×640     | $0.080 |       25%       |

### Product codes

| Model                 | Product                      |
| --------------------- | ---------------------------- |
| FLUX.1 Kontext \[dev] | `inference-flux-kontext-dev` |
| FLUX.1 Kontext \[pro] | `inference-flux-kontext-pro` |
| FLUX.1 Kontext \[max] | `inference-flux-kontext-max` |

## FLUX.2 Klein 4B

Applies to the [FLUX.2 Klein](https://docs.prodia.com/models/flux-2-klein/) 4B distilled job types:

- `inference.flux-2.klein.4b.txt2img.v1` and `inference.flux-2.klein.4b.img2img.v1`
- `inference.flux-2.klein.txt2img.v1` and `inference.flux-2.klein.img2img.v1` (the default aliases, currently routed to 4B distilled)

Priced per image by output resolution and the number of input images.

> note:
>
> These prices take effect on 2026-10-01. Until then the current price sheet applies.

### Price per image

| Inputs                    | Up to 1 MP | Up to 2 MP | Up to 3 MP | Over 3 MP |
| ------------------------- | ---------: | ---------: | ---------: | --------: |
| Text-to-image (no inputs) |    $0.0005 |    $0.0027 |    $0.0040 |   $0.0059 |
| 1 input image             |    $0.0010 |    $0.0027 |    $0.0040 |   $0.0059 |
| 2 input images            |    $0.0017 |    $0.0034 |    $0.0068 |   $0.0084 |
| 3 input images            |    $0.0025 |    $0.0065 |    $0.0078 |   $0.0084 |
| 4 input images            |    $0.0033 |    $0.0065 |    $0.0080 |   $0.0109 |
| 5 input images            |    $0.0045 |    $0.0068 |    $0.0102 |   $0.0109 |
| 6 input images            |    $0.0054 |    $0.0088 |    $0.0116 |   $0.0132 |
| 7 input images            |    $0.0068 |    $0.0092 |    $0.0130 |   $0.0145 |
| 8 input images            |    $0.0100 |    $0.0108 |    $0.0148 |   $0.0170 |

[FLUX.2 Klein 4B Pricing](https://docs.prodia.com/models/flux-2-klein-4b-pricing/) — Megapixel bucket thresholds, example sizes and product codes.

## Nano Banana

Applies to the [Nano Banana](https://docs.prodia.com/models/nano-banana/) and [Gemini 3](https://docs.prodia.com/models/gemini-3/) image job types:

- **Nano Banana (Gemini 2.5 Flash)** – `inference.nano-banana.txt2img.v2`, `inference.nano-banana.img2img.v2` and `inference.nano-banana.img2img.v1`
- **Nano Banana Pro (Gemini 3 Pro)** – `inference.gemini-3-pro.txt2img.v1` and `inference.gemini-3-pro.img2img.v1`
- **Nano Banana 2 (Gemini 3.1 Flash)** – `inference.gemini-3-1-flash.txt2img.v1` and `inference.gemini-3-1-flash.img2img.v1`

Priced per image by the `resolution` setting (`1K`, `2K` or `4K`). Aspect ratio, prompt and input images do not change the price, and image-to-image bills the same as text-to-image.

### Price per image

| Model                            | Resolution | Example sizes                  |  Price |
| -------------------------------- | ---------- | ------------------------------ | -----: |
| Nano Banana (Gemini 2.5 Flash)   | 1K         | 1024×1024, 1568×672, 672×1568  | $0.039 |
| Nano Banana Pro (Gemini 3 Pro)   | 1K or 2K   | 1024×1024, 1568×672, 2560×1440 | $0.150 |
| Nano Banana Pro (Gemini 3 Pro)   | 4K         | 3840×2160                      | $0.300 |
| Nano Banana 2 (Gemini 3.1 Flash) | 1K         | 1024×1024                      | $0.080 |
| Nano Banana 2 (Gemini 3.1 Flash) | 2K         | 2048×2048                      | $0.120 |
| Nano Banana 2 (Gemini 3.1 Flash) | 4K         | 3840×2160                      | $0.160 |

Nano Banana (Gemini 2.5 Flash) only produces 1K output. `resolution` defaults to `1K` when omitted.

### Product codes

| Model                            | Product                                                                                        |
| -------------------------------- | ---------------------------------------------------------------------------------------------- |
| Nano Banana (Gemini 2.5 Flash)   | `inference-nano-banana`                                                                        |
| Nano Banana Pro (Gemini 3 Pro)   | `inference-gemini-3-pro`, `inference-gemini-3-pro-4k`                                          |
| Nano Banana 2 (Gemini 3.1 Flash) | `inference-gemini-3-1-flash`, `inference-gemini-3-1-flash-2k`, `inference-gemini-3-1-flash-4k` |

## Qwen Image Edit Plus

Available in two tiers, chosen by job type:

- **Speed** – `inference.qwen.image-edit.plus.lightning.img2img.v2`
- **Quality** – `inference.qwen.image-edit.plus.img2img.v1` and `inference.qwen.image-edit.plus.lightning.img2img.v1`

Priced per generation by output resolution and whether one or several input images are supplied. Steps, prompt and seed do not change the price.

### Price per generation

| Tier    | Inputs                | Up to 1 MP | Up to 4 MP |
| ------- | --------------------- | ---------: | ---------: |
| Speed   | Single input image    |    $0.0085 |    $0.0320 |
| Speed   | Multiple input images |    $0.0170 |    $0.0650 |
| Quality | Single input image    |    $0.0150 |    $0.0450 |
| Quality | Multiple input images |    $0.0300 |    $0.0900 |

### How the buckets work

*Output resolution* is measured in pixels (`width × height`, each 256 to 2048):

| Bucket     | Pixel range | Example sizes                   |
| ---------- | ----------- | ------------------------------- |
| Up to 1 MP | ≤ 1,048,576 | 1024×1024, 1200×900, 1280×800   |
| Up to 4 MP | > 1,048,576 | 1600×1200, 1920×1080, 2048×2048 |

*Input images* are counted from the `images` array (1 to 3). Exactly one image bills as single input; two or three bill as multiple inputs.

### Product codes

The `product` field follows the pattern `inference-qwen-image-edit-{speed,quality}-{small,large}-{single,multi}`, where `small` is the up-to-1 MP bucket and `large` is the up-to-4 MP bucket. For example a Speed job with two input images rendered at 1920×1080 bills as `inference-qwen-image-edit-speed-large-multi`.

## Recraft

Applies to the Recraft V3 and [Recraft V4](https://docs.prodia.com/models/recraft-v4/) job types. Each is a flat price per generation; `size`, prompt and style do not change the price.

### Price per generation

| Model         | Output       | Job types                                                      | Price |
| ------------- | ------------ | -------------------------------------------------------------- | ----: |
| V3            | Raster image | `inference.recraft.txt2img.v1`, `inference.recraft.img2img.v1` | $0.04 |
| V3 Vectors    | SVG          | `inference.recraft.img2vec.v1`                                 | $0.08 |
| V4            | Raster image | `inference.recraft.v4.txt2img.v1`                              | $0.04 |
| V4 Pro        | Raster image | `inference.recraft.v4.pro.txt2img.v1`                          | $0.25 |
| V4 Vector     | SVG          | `inference.recraft.v4.txt2vec.v1`                              | $0.08 |
| V4 Pro Vector | SVG          | `inference.recraft.v4.pro.txt2vec.v1`                          | $0.30 |

V3 and V4 render up to about 1 MP (for example 1024×1024 or 2048×1024 for V3, 768×1536 or 1024×1024 for V4). V4 Pro renders up to about 4 MP (for example 2048×2048 or 1536×3072).

### Product codes

| Model         | Product                           |
| ------------- | --------------------------------- |
| V3            | `inference-recraft`               |
| V3 Vectors    | `inference-recraft-img2vec`       |
| V4            | `inference-recraft-v4`            |
| V4 Pro        | `inference-recraft-v4-pro`        |
| V4 Vector     | `inference-recraft-v4-vector`     |
| V4 Pro Vector | `inference-recraft-v4-pro-vector` |

## SDXL

Applies to the [SDXL](https://docs.prodia.com/models/sdxl/) job types `inference.sdxl.txt2img.v1`, `inference.sdxl.img2img.v1` and `inference.sdxl.inpainting.v1`.

Priced per image by step count. Output resolution (512 to 1536 per side), prompt and seed do not change the price, and image-to-image and inpainting bill the same as text-to-image.

### Price per image

| Steps        | Example sizes                 |   Price |
| ------------ | ----------------------------- | ------: |
| Up to 25     | 1024×1024, 768×1024, 1024×768 | $0.0020 |
| 26 to 50     | 1024×1024, 768×1024, 1024×768 | $0.0025 |
| 51 and above | 1024×1024, 768×1024, 1024×768 | $0.0100 |

### Product codes

The `product` field is `inference-sdxl-steps-25`, `inference-sdxl-steps-50` or `inference-sdxl-steps-100` for the three step buckets.

## Seedream

Applies to the [Seedream](https://docs.prodia.com/models/seedream/) job types:

- **Seedream 4.0** – `inference.seedream-4.txt2img.v1` and `inference.seedream-4.img2img.v1`
- **Seedream 4.5** – `inference.seedream-4-5.txt2img.v1` and `inference.seedream-4-5.img2img.v1`
- **Seedream 5.0 Lite** – `inference.seedream-5-0.lite.txt2img.v1` and `inference.seedream-5-0.lite.img2img.v1`

Flat price per image at any supported resolution. Output size, prompt, seed and input images do not change the price, and image-to-image bills the same as text-to-image.

### Price per image

| Model             | Example sizes                   |  Price |
| ----------------- | ------------------------------- | -----: |
| Seedream 4.0      | 1024×1024, 2048×2048, 4096×4096 | $0.030 |
| Seedream 4.5      | 1024×1024, 2048×2048, 4096×4096 | $0.040 |
| Seedream 5.0 Lite | 1024×1024, 2048×2048            | $0.035 |

### Product codes

| Model             | Product                       |
| ----------------- | ----------------------------- |
| Seedream 4.0      | `inference-seedream-4`        |
| Seedream 4.5      | `inference-seedream-4-5`      |
| Seedream 5.0 Lite | `inference-seedream-5-0-lite` |

## Image Utilities

Fixed-price utility jobs. Each bills a single flat product per call; input image size, prompt and other settings do not change the price. Volume pricing applies from 1M generations.

### Price per generation

| Utility              | Model                      | Job type                                 |   Price | Volume discount |
| -------------------- | -------------------------- | ---------------------------------------- | ------: | :-------------: |
| Background removal   | BiRefNet 2                 | `inference.remove-background.v1`         | $0.0025 |       25%       |
| NSFW image detection | ViT                        | `inference.vit.img2label.v2`             | $0.0002 |        —        |
| Upscale 2×           | R-ESRGAN                   | `inference.upscale.v1` with `upscale: 2` | $0.0010 |       25%       |
| Upscale 4×           | R-ESRGAN                   | `inference.upscale.v1` with `upscale: 4` | $0.0020 |       25%       |
| Upscale 8×           | R-ESRGAN                   | `inference.upscale.v1` with `upscale: 8` | $0.0030 |       25%       |
| Upscale 2×           | HYPIR                      | `inference.hypir.upscale.v1`             | $0.0500 |       25%       |
| Face restore         | GFPGAN                     | `inference.facerestore.v1`               | $0.0008 |       25%       |
| Segmentation         | SAM 3 (Segment Anything 3) | `inference.sam3.segment.v1`              | $0.0050 |       25%       |

### Product codes

| Utility              | Product                                                                |
| -------------------- | ---------------------------------------------------------------------- |
| Background removal   | `inference-background-removal`                                         |
| NSFW image detection | `inference-vit-nsfw`                                                   |
| Upscale (R-ESRGAN)   | `inference-upscale-2x`, `inference-upscale-4x`, `inference-upscale-8x` |
| Upscale (HYPIR)      | `inference-hypir-upscale-v1`                                           |
| Face restore         | `inference-facerestore`                                                |
| Segmentation         | `inference-sam3`                                                       |

[Job Pricing](https://docs.prodia.com/reference/price/) — Retrieve the exact cost of a job with the price query parameter.

[Tracking Job Costs](https://docs.prodia.com/guides/tracking-costs/) — Monitor spend across jobs and tokens.