Seedream 5.5 Uncensored by ByteDance
ByteDance Seedream 5.5 Uncensored creates or edits images from a prompt, supports transparent PNG backgrounds, and stays stable across long edit chains.
Overview
Seedream 5.5 Uncensored is an image generation and instruction-based editing model from ByteDance’s Seed team. It can generate a new image from your description or apply targeted changes to a provided image.
The model is tuned for multi-turn editing. It aims to keep unedited regions pixel-stable across many successive edits. It also supports native RGBA output, which makes real transparent-background assets without background removal.
What you can build
- Product cutouts, stickers, and icons with real transparency for design tools
- Multi-round art direction workflows where you refine one image step by step
- Marketing creatives with small, legible text and consistent layout
- Character and object assets you can place on new backgrounds without matting
- Before/after edits like outfit swaps, prop changes, relighting, and scene changes
Inputs
- A required written prompt that describes what to generate or how to change the image
- An optional single input image to edit; if you don’t provide one, it generates from the prompt
- A required output resolution choice (1K, 1.5K, or 2K)
- An optional aspect ratio choice (or auto), used to shape the final frame
- An optional background mode: automatic, fully opaque, or transparent
- An optional output file format choice: JPEG or PNG
- An optional watermark toggle that adds an “AI generated” mark in the corner
Outputs
The model returns 1 generated image file.
- If you choose JPEG, you get an opaque image.
- If you choose PNG, you can get either an opaque PNG or an RGBA PNG with an alpha channel.
- If you request a transparent background, the output includes real transparency, including semi-transparent edges.
- The image matches your selected resolution tier and aspect ratio choice.
Limitations
- Output size is constrained by a pixel-area budget. Extreme aspect ratios may be rejected or forced into a supported range.
- Transparent-background output requires PNG. If you request transparency with JPEG output, it won’t work.
- Multi-turn edits can still drift over time. Fine details may soften after many revisions.
- Small text is improved, but it can still misspell words or alter characters. Proofread any text.
- Hair, fur, smoke, lace, and glass edges can show halos in transparent PNG output. Inspect edges before production use.
- Low-quality inputs increase failure risk. Heavy JPEG artifacts, motion blur, screenshots with UI clutter, and scanned images can cause unwanted changes.
Safety & compliance
- “Uncensored” does not mean unmoderated. Provider-side and platform-side rules can still block some requests.
- Content involving minors, or anyone who could be under 18, is never allowed.
- Do not create sexual content involving minors, exploitative content, or non-consensual content.
- Don’t upload images you don’t have rights to use. You’re responsible for consent and permissions.
- Avoid requests that impersonate real people or create deceptive deepfakes, especially for sexual content or harassment.
Example prompts
Great starting points for seedream 5.5 uncensored.
API quick start
Run seedream 5.5 uncensored with a single API call.
{
"prompt": "Keep the same model, her face and blonde …",
"resolution": "1k",
"inputImage": "https://your-cdn.com/input.png",
"aspectRatio": "auto"
}curl -X POST "https://api.wiro.ai/v1/Run/bytedance/seedream-5-5-uncensored" \
-H "Content-Type: application/json" \
-H "x-api-key: YOUR_WIRO_API_KEY" \
--data-binary @- <<'JSON'
{
"prompt": "Keep the same model, her face and blonde …",
"resolution": "1k",
"inputImage": "https://your-cdn.com/input.png",
"aspectRatio": "auto"
}
JSONOther models you might like
Explore similar image generation models.



