bytedance/seedream-3 🔢❓📝 → 🖼️

⭐ Official ▶️ 2.7M runs 📅 Jun 2025 ⚙️ Cog 0.16.7 📄 Paper
2k-resolution chinese-english high-resolution text-rendering text-to-image

About

A text-to-image model with support for native high-resolution (2K) image generation

Example Output

Prompt:

"A cinematic, photorealistic medium shot capturing the nostalgic warmth of a mid-2000s indie film. The focus is a young woman with a sleek, straight bob haircut in cool platinum white with freckled skin, looking directly and intently into the camera lens with a knowing smirk, her head is looking up slightly. She wears an oversized band t-shirt that says "Seedream 3.0 on Replicate" in huge stylized text over a long-sleeved striped top and simple silver stud earrings. The lighting is soft, golden hour sunlight creating lens flare and illuminating dust motes in the air. The background shows a blurred outdoor urban setting with graffiti-covered walls (the graffiti says "seedream" in stylized graffiti lettering), rendered with a shallow depth of field. Natural film grain, a warm, slightly muted color palette, and sharp focus on her expressive eyes enhance the intimate, authentic feel"

Output

Example output

Performance Metrics

5.66s Prediction Time
5.68s Total Time
All Input Parameters
{
  "size": "regular",
  "width": 2048,
  "height": 2048,
  "prompt": "A cinematic, photorealistic medium shot capturing the nostalgic warmth of a mid-2000s indie film. The focus is a young woman with a sleek, straight bob haircut in cool platinum white with freckled skin, looking directly and intently into the camera lens with a knowing smirk, her head is looking up slightly. She wears an oversized band t-shirt that says \"Seedream 3.0 on Replicate\" in huge stylized text over a long-sleeved striped top and simple silver stud earrings. The lighting is soft, golden hour sunlight creating lens flare and illuminating dust motes in the air. The background shows a blurred outdoor urban setting with graffiti-covered walls (the graffiti says \"seedream\" in stylized graffiti lettering), rendered with a shallow depth of field. Natural film grain, a warm, slightly muted color palette, and sharp focus on her expressive eyes enhance the intimate, authentic feel",
  "aspect_ratio": "16:9",
  "guidance_scale": 2.5
}
Input Parameters
seed Type: integer
Random seed. Set for reproducible generation
size Default: regular
Big images will have their longest dimension be 2048px. Small images will have their shortest dimension be 512px. Regular images will always be 1 megapixel. Ignored if aspect ratio is custom.
width Type: integerDefault: 2048Range: 512 - 2048
Image width
height Type: integerDefault: 2048Range: 512 - 2048
Image height
prompt (required) Type: string
Text prompt for image generation
aspect_ratio Default: 16:9
Image aspect ratio. Set to 'custom' to specify width and height.
guidance_scale Type: numberDefault: 2.5Range: 1 - 10
Prompt adherence. Higher = more literal.
Output Schema

Output

Type: stringFormat: uri

Example Execution Logs
Using seed: 65795849
Generating 1280x720 image
Generating...
Generated in 5.7sec
Version Details
Version ID
745eb069a311b3a5c4f965934bf8240f37e883a4672ab49f2519bc02138c34d7
Version Created
October 1, 2025
Run on Replicate →