SeeDream - AI Image Generator
Explore API MODEL 
Choose an AI palette for artistic style.
Seedream 5 Lite
Transform your images.
Choose an AI palette for artistic style.
Transform your images.
SeeDream is an image-generation family built by the Seed Team at ByteDance. In the Pixazo playground it comes as four models: image modification from references (5 Lite), photorealistic text-to-image (4.5 and 3), and long-prompt text-to-image (4). Same greenhouse, four different seeds.
Upload up to 10 reference images (JPG/PNG, <25MB) as the visual base, then describe the modification in ≤1,000 characters — style, lighting, background, mood, realism, colors or composition.
Describe subject, environment, lighting and style in ≤1,000 characters; pick Portrait, Square or Landscape.
Up to 3,000 characters of description — subject, style, lighting, colors, mood, background and fine detail all steer the result.
3,000-character descriptions with camera-level detail — “cinematic portrait in the rain, neon reflections, shallow depth of field”.
Up to 10 JPG/PNG files under 25MB each — they establish the visual foundation SeeDream will modify.
Say exactly how the references should change: style, lighting, atmosphere, composition, colors, mood.
SeeDream applies the requested transformation; regenerate to refine the result.
With clear lighting descriptions — the models’ home turf.
Specify time of day and weather for dependable, dramatic results.
Simple, clean backgrounds keep the subject crisp and commercial.
The modification model follows your uploads — weak or conflicting references produce muddled output regardless of the prompt.
Lettering inside generated art can misspell; add critical text in an editor.
The 3,000-character models reward long, specific prompts — one-liners under-use them.
The same seed prompt yields variations, not clones. Save the bloom you like.
Four models, one greenhouse — describe the image and watch it bloom.
Generate with SeeDream — free