Seedream 5.5 Text-to-Image Spicy

seedream-5.5/text-to-image-global
SpicyText-to-Image

Seedream 5.5 Spicy is based on ByteDance Seed's Seedream 5.5 image generation model, in its Spicy mode (uncensored improved version) with relaxed content restrictions compared to the standard version. As a channel-customized variant that preserves the original base model's powerful generative engine, users must comply with applicable local laws and regulations. It offers 1K, 1.5K, and 2K resolutions (with 2K as default), supports 1–10 reference images, and enables up to 100 rounds of pixel-level multi-turn editing without color drift or detail loss. It natively generates RGBA transparent backgrounds and delivers accurate text rendering across 15+ languages.

Pricing Note: The prices shown on this page are iCreat AI’s standard rates. Some models may have limited-time discounts or other promotions, so actual charges may differ from the listed prices. Final charges are based on the real-time pricing displayed on the corresponding model’s Playground page.

Resolution Price (USD) Billing Unit
1K 0.02 Per image
1.5K 0.02 Per image
2K 0.035 Per image

Additional reference image fee: $0.008 per image.

Total cost = Price for the selected resolution + Number of reference images × $0.008.

Read Me

Seedream 5.5 Text-to-Image Spicy API

Seedream 5.5 Text-to-Image Spicy is based on ByteDance's Seedream 5.5 image generation model, in its Spicy mode (uncensored improved version) with relaxed content restrictions compared to the standard version.

On the iCreat platform the model is exposed through the asynchronous task endpoint https://api.icreat.ai/v1/task/submit/seedream-5.5/text-to-image-global, followed by result polling at https://api.icreat.ai/v1/task/result. It generates images directly from a text prompt with no reference image required, supports 1K / 1.5K / 2K resolution tiers (default 2048x2048, no 4K), and is billed per image at $0.02 for 1K and 1.5K and $0.035 for 2K, plus $0.008 per additional reference image.

Model Positioning

Seedream 5.5 Text-to-Image Spicy is a pure text-to-image generation model. It takes a natural-language prompt and returns a finished image, with no input image needed for the primary workflow.

Spicy is a channel naming convention created by the iCreat platform — the model capability comes from the original Seedream 5.5 base model, and Spicy is the platform's name for its uncensored variant. It is not an official Spicy version released by ByteDance. The platform tags this channel with the SPICY version tag.

The model sits at the high-fidelity end of the Seedream family: up to 2K output, native RGBA transparent-background PNG, and high-accuracy text rendering across 15+ languages. It is aimed at production workloads such as e-commerce marketing, film concept design, game art, and commercial advertising.

Core Capabilities

Pure Text-to-Image Generation

The endpoint accepts a single text prompt and returns a generated image. No reference image, mask, or source image is required, which makes it suitable for batch ideation, prompt-driven asset production, and fully automated pipelines.

2K High-Resolution Output

The model supports 1K, 1.5K, and 2K resolution tiers, with a default size of 2048x2048. Output quality is sufficient for commercial delivery at the 2K tier, and the lower tiers provide a cheaper path for drafts and rapid iteration. 4K is not supported.

Native RGBA Transparency

The model can produce native RGBA transparent-background PNG output, so cut-out assets can be generated directly without a separate background-removal step. This is useful for product shots, stickers, overlays, and compositing workflows.

15+ Language Text Rendering

Text rendering inside the generated image is high-accuracy and covers 15+ languages, which allows posters, packaging mockups, and advertising creatives to carry legible typography without post-editing.

Multi-Resolution and Multi-Aspect-Ratio

28 sizes are supported across the three tiers, covering aspect ratios of 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, and 21:9. The 2K tier retains the legacy official sizes, so existing prompt configurations remain compatible.

Optional Reference Image Guidance

Although the primary use of this endpoint is pure text-to-image, it optionally accepts 1–10 reference images to preserve subject, product, or style consistency. Each reference image adds $0.008 to the cost of the request. The reference-image field is named image.

Asynchronous Two-Step Task API

Generation runs as an asynchronous task: submit a job to the submit endpoint, then poll the task result endpoint until the task reaches a terminal state. This fits long-running batch generation and server-side job queues.

Pricing

Resolution Price (USD) Billing Unit
1K $0.02 Per image
1.5K $0.02 Per image
2K $0.035 Per image

Additional reference image fee: $0.008 per image.

Total cost = Price for the selected resolution + Number of reference images × $0.008.

Live-verified billing examples: 1K = 0.0200; 1.5K = 0.0200; 2K = 0.0350; 2K with 1 reference image = 0.0430; 1K with 3 reference images = 0.0440.

Note: billing is per generated image, not per token or per second. The reference-image surcharge applies only when the image field is actually supplied and accepted; a request that sends the wrong field name is not charged the surcharge. Prices listed here reflect iCreat AI's standard rates. Some models may be eligible for promotional discounts or special offers. Final charges are based on the real-time prices displayed in each model's Playground.

Application Scenarios

  • E-commerce marketing: product hero images, promotional banners, and catalog visuals generated from copy.
  • Film and series concept design: mood boards, key art, and scene exploration from script descriptions.
  • Game art: character concepts, environment keyframes, item icons, and UI illustration assets.
  • Commercial advertising: campaign creatives, poster layouts, and multi-language typography mockups.
  • Transparent-background asset production: cut-out PNG assets for compositing and design systems.
  • Batch prompt-driven content pipelines that need consistent output at a fixed resolution tier.

Model Comparison

Table 1 — Seedream 5.5 Text-to-Image Spicy vs Seedream 5.5 Text-to-Image (standard)

Item Spicy Standard
Base model Seedream 5.5 Seedream 5.5
Capability type Text-to-image Text-to-image
Resolution tiers 1K / 1.5K / 2K 1K / 1.5K / 2K
Reference image limit 10 10
Prompt limit 10,000 characters 10,000 characters
Billing $0.02 (1K) / $0.02 (1.5K) / $0.035 (2K) per image, +$0.008 per reference image $0.02 (1K) / $0.02 (1.5K) / $0.035 (2K) per image, +$0.008 per reference image
Content restriction level Relaxed (uncensored variant) Standard

Note: the two channels share an identical base model, identical resolution tiers, and identical billing. The only difference is the content restriction level. Spicy is an iCreat platform channel naming convention, not a separate official model release.

Table 2 — Latest comparable image models from the same developer/class

Item Seedream 5.5 Text-to-Image Spicy Seedream 5.5 Edit Spicy Seedream 5.0 Pro
Capability type Text-to-image Image-to-image editing + text-to-image Text-to-image
Resolution tiers 1K / 1.5K / 2K 1K / 1.5K / 2K 2K only (seven sizes)
Unit price 1K $0.02 / 1.5K $0.02 / 2K $0.035 1K $0.02 / 1.5K $0.02 / 2K $0.035 $0.09 per image
Reference image limit 10 10 10
Reference image surcharge $0.008 per image $0.008 per image Not applicable

Note: Seedream 5.5 Edit Spicy is an editing-oriented sibling that also accepts text-to-image requests; Seedream 5.0 Pro is the previous-generation high-resolution model priced at $0.09 per image with a 2K-only size set and a 10,000-character prompt limit. Billing basis may differ between generations — always confirm the real-time price in each model's Playground.

Why Choose Seedream 5.5 Text-to-Image Spicy?

  • Pure text-to-image generation with no source image required — ideal for prompt-driven and automated pipelines.
  • Up to 2K output at $0.035 per image, with 1K and 1.5K tiers at $0.02 for cheaper drafts.
  • Native RGBA transparent-background PNG, removing a separate cut-out step.
  • High-accuracy text rendering across 15+ languages for posters, packaging, and ad creatives.
  • 28 verified sizes across 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, and 21:9.
  • Optional 1–10 reference images for subject, product, or style consistency at $0.008 each.
  • Relaxed content restrictions relative to the standard channel, via the iCreat SPICY channel tag.
  • Asynchronous two-step task API that fits server-side batch and queue-based workloads.

Specifications

Field Value
Model Name Seedream 5.5 Text-to-Image Spicy
Developer ByteDance (Seed team)
Model Code seedream-5.5/text-to-image-global
Endpoint https://api.icreat.ai/v1/task/submit/seedream-5.5/text-to-image-global
Task Result Endpoint https://api.icreat.ai/v1/task/result
Release Date 2026-11-09
iCreat Listing Date 2026-11-12
Model Type Text-to-image (textToImage)
Authentication Authorization: Bearer ${ICREAT_API_KEY}
Version Tag SPICY
Input Modalities Text; optional reference images
Output Modalities Image (PNG or JPEG)
Supported Resolutions 1K / 1.5K / 2K
Aspect Ratios 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3, 21:9
Reference Image Limit 10 (max 30 MB each)
Reference Image Surcharge $0.008 per image
Prompt Limit 10,000 characters
Default Size 2048x2048
Watermark Default true; pass false to disable
Output Format png (default) or jpeg
Billing Unit Per image
API Mode Asynchronous two-step (submit task → poll result)

input fields

Field Type Required Default Description
prompt string Yes — Text instruction describing the image to generate. Maximum 10,000 characters.
size string Yes — Output size in WIDTHxHEIGHT form. Must be one of the 28 supported values. Omitting it returns HTTP 400 size is required.
image string[] No — Optional reference images (1–10 URLs, max 30 MB each) used to preserve subject, product, or style consistency. Each image adds $0.008. The field name is image, not images.

parameters fields

Field Type Required Default Description
watermark boolean No true Whether to apply a watermark. Watermark is on by default; pass false explicitly to disable.
output_format string No png Output file format. Accepts png and jpeg. jpg is invalid and returns HTTP 400 output_format must be png or jpeg.

Supported sizes by tier

1K (8): 1024x1024, 1152x864, 864x1152, 1424x800, 800x1424, 1248x832, 832x1248, 1568x672

1.5K (8): 1536x1536, 1792x1344, 1344x1792, 2048x1152, 1152x2048, 1872x1248, 1248x1872, 2352x1008

2K (12): 2048x2048, 2368x1776, 1776x2368, 2816x1584, 1584x2816, 2496x1664, 1664x2496, 3136x1344, 2304x1728, 1728x2304, 2720x1530, 1530x2720

Note: the 2K tier retains the legacy official sizes (2304x1728, 1728x2304, 2720x1530, 1530x2720), all verified working.

Architecture

Seedream 5.5 is a diffusion-based image generation model developed by ByteDance's Seed team. It maps a text prompt (and, optionally, reference images) into the latent space and decodes a final image at the requested resolution. The iCreat platform exposes it through an asynchronous task API: the submit endpoint enqueues a generation job and returns a task identifier, and the task result endpoint reports status and, on success, the generated image URL.

Generation is performed at the requested size tier rather than being upscaled afterwards, which is why the 1K, 1.5K, and 2K tiers are billed differently. Native RGBA output and multi-language text rendering are handled inside the generation pipeline, so transparent backgrounds and in-image typography do not require post-processing.

Because the workload is asynchronous, transient upstream congestion can surface as a terminal failure. In that case the task returns {"status":"FAILED","error_code":"PROVIDER-BUSY","error_message":"PROVIDER-BUSY","result":[]}, which indicates temporary upstream congestion and is safe to retry.

Notes

  • The reference-image field is image (singular, string[]), not images. Sending images is silently ignored by the gateway — no error is raised, an image is still returned, but the reference image does not take effect and the $0.008 surcharge is not charged. Always use image.
  • output_format accepts only png and jpeg. The value jpg is invalid and returns HTTP 400 with output_format must be png or jpeg.
  • Only prompt and size are hard-required. Omitting size returns HTTP 400 size is required; omitting prompt returns HTTP 400 Invalid request body.
  • watermark may be omitted; it defaults to true, meaning the watermark is on by default. Pass false explicitly to disable it.
  • A maximum of 10 reference images is accepted. Sending 11 returns HTTP 400 image must contain at most 10 URLs. Each reference image must be at most 30 MB.
  • The prompt limit is 10,000 characters.
  • 4K output is not supported. The maximum resolution tier is 2K.
  • If a task terminates with PROVIDER-BUSY, this indicates transient upstream congestion and the request is safe to retry.
  • Spicy is a channel naming convention created by the iCreat platform. The model capability comes from the original Seedream 5.5 base model; Spicy is the platform's name for its uncensored variant and is not an official ByteDance release. The platform tags this channel with the SPICY version tag.
  • Billing formula: Total cost = Price for the selected resolution + Number of reference images × $0.008.

Frequently Asked Questions

How do I call Seedream 5.5 Text-to-Image Spicy?

Submit a POST request to https://api.icreat.ai/v1/task/submit/seedream-5.5/text-to-image-global with the header Authorization: Bearer ${ICREAT_API_KEY} and a JSON body containing at least prompt and size. The response returns a task identifier; poll https://api.icreat.ai/v1/task/result with that task identifier until the task reaches a terminal state, then read the generated image from the result.

How is billing calculated?

Billing is per generated image, based on the selected resolution tier: $0.02 for 1K, $0.02 for 1.5K, and $0.035 for 2K. If reference images are supplied through the image field, each one adds $0.008. Total cost = Price for the selected resolution + Number of reference images × $0.008. For example, a 2K image with 1 reference image costs $0.0430, and a 1K image with 3 reference images costs $0.0440.

Is a reference image required?

No. This is a text-to-image endpoint — the primary workflow uses only prompt and size. Reference images are an optional capability used to preserve subject, product, or style consistency, and they add $0.008 per image.

Which field name should I use for reference images?

Use image (singular, string[]). Do not use images. Sending images is silently ignored by the gateway: no error is returned, an image is still generated, but the reference image does not take effect and the $0.008 surcharge is not charged. This silent failure makes the wrong field name easy to miss, so always verify you are sending image.

What values does output_format accept?

It accepts png and jpeg, and defaults to png. It may be omitted entirely. The value jpg is invalid and returns HTTP 400 with output_format must be png or jpeg. Note that requesting jpeg produces a .jpg file.

Which sizes are supported?

28 sizes are supported across three tiers. 1K (8): 1024x1024, 1152x864, 864x1152, 1424x800, 800x1424, 1248x832, 832x1248, 1568x672. 1.5K (8): 1536x1536, 1792x1344, 1344x1792, 2048x1152, 1152x2048, 1872x1248, 1248x1872, 2352x1008. 2K (12): 2048x2048, 2368x1776, 1776x2368, 2816x1584, 1584x2816, 2496x1664, 1664x2496, 3136x1344, 2304x1728, 1728x2304, 2720x1530, 1530x2720. The default is 2048x2048.

What is the default behavior of watermark?

watermark may be omitted, and it defaults to true — the watermark is on by default. Pass false explicitly if you need a clean image.

What is the difference between Spicy and the standard channel?

Both channels share the same base model, the same resolution tiers, and the same billing. The only difference is the content restriction level: Spicy is the uncensored variant with relaxed content restrictions compared to the standard version.

Is Spicy an official ByteDance release?

No. Spicy is a channel naming convention created by the iCreat platform. The model capability comes from the original Seedream 5.5 base model, and Spicy is the platform's name for its uncensored variant. It is not an official Spicy version released by ByteDance. The platform tags this channel with the SPICY version tag.

How does the asynchronous flow work, and what happens if a task fails?

The flow is two-step: submit the job to https://api.icreat.ai/v1/task/submit/seedream-5.5/text-to-image-global, then poll https://api.icreat.ai/v1/task/result with the returned task identifier until the task reaches a terminal state. If a task terminates with {"status":"FAILED","error_code":"PROVIDER-BUSY","error_message":"PROVIDER-BUSY","result":[]}, this indicates transient upstream congestion and the request is safe to retry.