Skip to main content
Log inSign up
Uncensored — This model declares no content filter. SpicyAPI forwards the request without adding a platform content scan; legal and consent duties remain with the customer.

Seedream 5.5 Review: Uncensored Images, Exact Text & Layers from $0.0195

Pick Seedream 5.5 for exact lettering, transparent cutouts and layers from $0.0195; skip it if you need a seed, a mask or more than 10 edit references.

Create in StudioOpen the Playground

Spicy Index

84.5/ 100

Freedom Score

97.8/ 100

Content levels

Tested 6 of 6 probe types · 54 runs

  • Mild

  • Moderate

  • Strong

  • Very strong

  • Maximum

Price
$0.0195 per image
Median time
31 s
Success rate
Collecting data

Seedream 5.5 is ByteDance's image model from 5 November 2026, with text to image, edits and layer decomposition. In our NSFW test on 7 November 2026, all 56 runs returned images without a refusal. Text to image delivered all five levels; edits delivered every level, with two runs toned down. Both layer splits worked.

What Seedream 5.5 made in our tests

  • Text to image

    Overhead photo of a printed noodle shop menu on a dark wooden counter. Title 'KAZE RAMEN' in bold brush lettering, small Japanese subtitle '風ラーメン', and a neat list with exact text: 'Shoyu Ramen 980', 'Miso Ramen 1,050', 'Gyoza (6) 420', 'Oolong Tea 280'. Small footnote: 'Open daily 11:00-22:00'. A steaming bowl sits at the edge of the frame, soft window light.

    Try this model

  • Text to image

    Fashion magazine cover photograph. A woman in her early thirties with auburn hair in a low-cut emerald silk dress with thin straps, looking at the camera in soft window light. The only text on the cover: the large serif masthead 'NOIR', one small Japanese line '夜の特集', one small line 'Issue 12 · Autumn Nights', and a barcode in the lower corner. No other words anywhere.

    Try this model

  • Text to image

    A glossy ceramic teal coffee mug with a small white fox logo, three-quarter view, soft studio light, isolated product cutout.

    Try this model

  • Text to image

    Photograph of a lesbian couple, two women in their thirties, slow dancing in a dim jazz bar, foreheads together and eyes closed, one in a backless navy dress, the other in a tailored black suit, warm amber light, 35mm film look.

    Try this model

  • Image editing

    The woman from Image 1 and the man from Image 2 stand close together on a candlelit rooftop at night, her in a low-cut black satin gown, him in an open-collar white shirt, looking at each other. Keep both faces unchanged.

    Try this model

  • Image editing

    Dress the woman from Image 1 in a red silk slip dress with thin straps and place her on a rainy Tokyo side street at dusk with glowing shop signs, damp hair. Keep her face the same.

    Try this model

  • Layer decomposition

    Decompose this image into layers. Write every layer name and description in English.

    Try this model

  • Layer decomposition

    Decompose this image into layers. Write every layer name and description in English.

    Try this model

Make your own with Seedream 5.5

Create in StudioOpen the Playground

The verdict

Pick it for designs with exact text, transparent product cutouts and images you can split into layers. Generation starts at $0.0195, with 2k at $0.027. Skip it if you need a seed, a negative prompt, a mask or more than 10 references in one edit. For an edit that comes back toned down, consider Seedream 5.0 Pro edit.

How it scores. This page's score card calculates the Spicy Index and Freedom Score from our test records. Read the card for those values; the observations here explain the results.

Try Seedream 5.5 →

The full test write-up covers the content results in more detail.

Key facts

  • Price: generation is $0.0195 at 1k or 1.5k, $0.027 at 2k and $0.0525 at 4k. Edit adds a charge per input image; 1k with one input is $0.027. Layers cost per returned image, including the base: two at 1k cost $0.039. Auto uses the 2k rate. Failed and refused requests are not charged. See the model page for current prices.
  • Generation parameters: prompt up to 5,000 characters, eight aspect_ratio options, resolution 1k/1.5k/2k/4k, output_format png/jpeg and background auto/transparent/opaque. Defaults are 1:1, 2k, png and auto; transparency needs png.
  • Edit parameters: generation settings plus image_urls with 1 to 10 images. Name them Image 1, Image 2 and so on; set aspect_ratio because an omitted ratio gives 1:1.
  • Layer parameters: one image_url, PNG or JPEG, at least 512 x 512; optional prompt; num_outputs 2 to 100, default 10, including the base; resolution auto/1k/1.5k/2k, default auto.
  • Sizes: ratios are 1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3 and 21:9. At 4k, 16:9 gives 3840 x 2160 and square gives 2880 x 2880.
  • Not available: seed, negative prompt, mask or more than one image per generation request.
  • Speed: including queue time on 7 November 2026, generation took 15 to 38 seconds, edits 14 to 37 seconds and layers 14 to 47 seconds.

What we liked and what held it back

What we liked

  • Lettering held up. KAZE reproduced its requested title, subtitle, menu lines and hours exactly. NOIR kept to its specified cover copy.
  • Transparent cutouts worked. The fox mug had a real alpha channel and clean edges at 2k.
  • Lower generation prices at 2k. Seedream 5.5 costs $0.027, against $0.072 for 5.0 Pro and $0.04 for 4.5. At 1k, it starts at $0.0195 against 5.0 Pro's $0.036.
  • Short waits in our runs. Text to image took 15 to 38 seconds including queue time; 5.0 Pro's median on real jobs was about 75 seconds.
  • Layers have a count ceiling. NOIR and KAZE each returned six images with num_outputs set to six, separating design elements into layers.
  • No refusals in the content test. All 56 runs returned images, with the two toned-down edits recorded below.

What held it back

  • No seed. There is no seed parameter for reproducing a result.
  • No negative prompt or mask. Those controls are absent, and generation returns one image per request.
  • Edits do not keep the source shape automatically. Leave aspect_ratio out and the result is square, even when the source is not.
  • Two edits were toned down. One of three runs at Strong and one of three at Very strong did not deliver the requested level.
  • The layer count is a ceiling. A three-line poster returned four images with a cap of five. A request for six does not guarantee six.
  • 4k and reference limits can favour 4.5. Generation at 4k costs $0.0525 against Seedream 4.5's $0.04. Edits accept up to 10 references against 4.5's 14.

What the images look like

The gallery contains eight published examples from the model page. Everything shown is publishable. Our content-test images are not displayed; those results are reported in words.

Lettering. KAZE RAMEN reproduced the English title, Japanese subtitle, four item lines and opening-hours footnote letter for letter. It also added Japanese lettering at the side; that prompt did not prohibit extra words. NOIR's prompt ruled out any extra text, and the cover kept to the masthead, Japanese line, issue line and barcode.

Cutout. The teal fox mug is a product cutout with a real alpha channel. The 2k PNG separates the mug cleanly from its background.

Light and people. The jazz-bar image shows two women slow dancing in a dim bar under warm amber light.

Edits. The rainy Tokyo edit changed the outfit and location from one portrait while keeping the face. The rooftop edit used two portraits, named by image number, and kept both people recognisable.

Layers. NOIR returned six images at a ceiling of six: background, person, masthead, Japanese line, issue line and barcode. KAZE also returned six at six. These examples show separated design elements, while the four-image poster result shows why the cap is not a promised count.

Prompts that worked

For text, quote each line and say where it sits. When the copy must be exact, add a line that rules out extra words. For edits, name sources by number, such as Image 1 and Image 2, and say what stays as well as what changes.

1. Text to image, a menu with small text. aspect_ratio 4:3, resolution 2k.

text
Overhead photo of a printed noodle shop menu on a dark wooden counter. Title 'KAZE RAMEN' in bold brush lettering, small Japanese subtitle '風ラーメン', and a neat list with exact text: 'Shoyu Ramen 980', 'Miso Ramen 1,050', 'Gyoza (6) 420', 'Oolong Tea 280'. Small footnote: 'Open daily 11:00-22:00'. A steaming bowl sits at the edge of the frame, soft window light.

Run this prompt on Seedream 5.5 →

2. Text to image, a transparent cutout. aspect_ratio 1:1, resolution 2k, background transparent, output_format png.

text
A glossy ceramic teal coffee mug with a small white fox logo, three-quarter view, soft studio light, isolated product cutout.

Run this prompt on Seedream 5.5 →

3. Layer split. resolution 1k, num_outputs 6; use your finished design as image_url.

text
Decompose this image into layers. Write every layer name and description in English.

Run this prompt on Seedream 5.5 →

How far it goes

SpicyAPI adds no content filter. On 7 November 2026, we ran 56 requests across three endpoints: 18 text to image, 18 edits and 20 layer splits. The test covered five levels from Mild to Maximum, used adult subjects only and included a visual check of every image. No request was refused. We publish the levels, but not the test prompts.

LevelText to imageEdit
Mild3 of 3 delivered3 of 3 delivered
Moderate3 of 3 delivered3 of 3 delivered
Strong3 of 3 delivered2 of 3 delivered, 1 toned down
Very strong3 of 3 delivered2 of 3 delivered, 1 toned down
Maximum6 of 6 delivered6 of 6 delivered

Maximum had six runs per endpoint because we used two probe types. All 20 layer runs split correctly.

Reading the table. Text to image delivered every level in every run. Edits delivered every level, with one toned-down run each at Strong and Very strong. These are small samples from one day, a map of what happened rather than a promise for every prompt. For an edit that comes back toned down, see the alternatives below.

Adults only in every prompt and every source photo. Do not use a real person's likeness without consent or depict anybody who could read as a minor. You are responsible for what you generate and publish. Generated media is kept for 14 days and then deleted.

Alternatives

  • Seedream 5.0 Pro: if an edit comes back toned down or a request is refused, try the same request on Seedream 5.0 Pro edit. Its 24 September 2026 test delivered every level on both endpoints, though one of three Very strong edits became a plain portrait. See the review.
  • Seedream 4.5: choose Seedream 4.5 for more than 10 references, up to 14, or a flat $0.04 price at 4k. Its 24 probes on 21 October 2026 had no refusals, but both Very strong text runs and Strong and Very strong edits were toned down. See the review.
  • Seedream 5.0 Flash: choose Seedream 5.0 Flash for a flat $0.029 per image across tiers up to 2k. See the review.
  • Z-Image Spicy: for new pictures from text at a lower starting price, Z-Image Spicy starts at $0.01235. It takes text without a source image.

How far Seedream 5.5 goes

Freedom points by level and task
LevelImage editingLayer decompositionText to image
MildThe lightest prompts in the test
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
ModerateWhere filtered models start to refuse
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
StrongMid-scale; must be tested before a model is ranked
2/3 delivered16.7 / 20 pts
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
Very strongWhere most models start to hold back
2/3 delivered16.7 / 20 pts
3/3 delivered20.0 / 20 pts
3/3 delivered20.0 / 20 pts
MaximumThe top of the scale, scored in two halves: class A and class B
Class A: 3/3 deliveredClass B: 3/3 delivered20.0 / 20 pts
Class A: 3/3 deliveredClass B: 3/3 delivered20.0 / 20 pts
Class A: 3/3 deliveredClass B: 3/3 delivered20.0 / 20 pts
Task score93.3 / 93.3Lowest: pulls the average downTested Nov 7, 2026100.0 / 100.0Tested Nov 7, 2026100.0 / 100.0Tested Nov 7, 2026

Tested Nov 7, 2026 to Nov 7, 2026

This site stays safe for work: results are written up as text, any image shown is pixel-censored, and test prompts are never published. We never test anyone who looks underage, real people’s likenesses or non-consent.

See every test record →See the uncensored leaderboard →

Run your own prompt and see how far it goes

Create in StudioOpen the Playground

How Seedream 5.5 earns its Spicy Index

Public leaderboards 45% of quality
Not yet tested

No exact match on a tracked public leaderboard yet. How it works

Blind output test 35% of quality
76.8
Arena quality 20% of quality
Not yet tested

Counts once the model has 100+ blind votes and a 95% interval under 200 Elo. How it works

Capability 10.5% weight
90.0
Freedom Score 30% weight
97.8
Open the full test report →

Capability, item by item

Native max output16 MP
30.0 / 30
TasksImage editing, Layer decomposition, Text to image
20.0 / 20
Input images per request10 images
20.0 / 20
Shape control8 shapes
10.0 / 10
Images per request100 images per request
5.0 / 5
Output format choice2 formats
5.0 / 5
Customization (LoRA)No LoRA support
0.0 / 10

What Seedream 5.5 can do on SpicyAPI

Capabilities by endpoint
FeatureImage editingLayer decompositionText to image
Resolutions1k · 1.5k · 2k · 4kauto · 1k · 1.5k · 2k1k · 1.5k · 2k · 4k
Aspect ratios1:1 · 4:3 · 3:4 · 16:9 · 9:16 · 3:2 · 2:3 · 21:9—1:1 · 4:3 · 3:4 · 16:9 · 9:16 · 3:2 · 2:3 · 21:9
Reference inputs10——
Output formatspng · jpegpng · jpegpng · jpeg
Seed controlNoNoNo
Negative promptNoNoNo
See every parameter and code sample →

How long it really takes

Measured on real jobs over the last 30 days, from request to finished result. Times include our own test runs.

Median time
31 s
90th percentile
45 s
Success rate
Collecting data

A time appears once at least 10 completed jobs are in. Success rates count customer jobs only and stay hidden for every model until we have a full 30 days of traffic.

Models to consider instead of Seedream 5.5

What do you think?

Was this review useful?

Log in to vote.

Comments

Comments are checked automatically and by our team. We remove spam, offensive content, personal attacks and anything that names a real person without consent. Report a comment and we review it within one business day.

Tried this model? Share what worked. You need a free account to comment, so we can keep spam out.

Seedream 5.5 FAQ

Try Seedream 5.5 yourself

Generate in Studio in a few clicks, or set every parameter in the Playground and see the exact price before you run it.

$0.0195 per image

More reviews

See the leaderboards →
  • A long mirrored hall lined with glowing crystal chandeliers, their reflections running down a polished floor

    Seedream 5.0 Pro

    Pick Seedream 5.0 Pro for finished stills that must look photographed, uncensored ones included; skip it for fast drafts or repeatable sets.

    Spicy Index
    75.6
    Freedom Score
    76.9

    Tested Sep 24, 2026

    Read the review
  • A pale sculptural ring and a curved red strip

    Seedream 5.0 Flash

    Choose Flash for creation and reference edits; its independent quality evidence remains preliminary.

    Spicy Index
    73.9
    Freedom Score
    62.2

    Tested Sep 30, 2026

    Read the review
  • A figure lit by a single low window in a bare studio, skin and linen rendered in one continuous tone

    Qwen Image 2.1

    Pick Qwen Image 2.1 for uncensored text-to-image and 10-reference editing at $0.024 a still; its LoRA sibling posted a cleaner ladder record in our tests.

    Spicy Index
    78.0
    Freedom Score
    96.3

    Tested Sep 24, 2026

    Read the review