Dreamina 3.1 is ByteDance's high-resolution image model — 4 megapixel output at 2K, cinematic aesthetic defaults, six aspect ratios including custom dimensions, and atmospheric scene composition that reads like film stills.
Dreamina 3.1 is ByteDance's high-resolution image model with cinematic aesthetic defaults — atmospheric lighting, painterly composition, and strong material rendering at the 4 megapixel (2K) tier. It's the model to reach for when you want output that already looks edited.
Six aspect ratio presets (1:1, 16:9, 9:16, 4:3, 3:4, plus 'custom' for specific dimensions 256-2048px), two resolution tiers (1K for iteration, 2K for hero shots), and an optional prompt-enhance toggle. Pricing is flat across resolutions, so prefer 2K unless you specifically need 1K speed.
4MP
2K resolution output
6
Aspect ratios + custom
Flat
Same price 1K vs 2K
Dreamina 3.1 produces cinematic-looking output from sparse prompts — atmospheric lighting, painterly color grading, strong depth. Other models often need explicit cinematography language to hit this look; Dreamina hits it by default.
2K (4 megapixel) output at the same credit cost as 1K — no reason not to default to it. Strong texture detail, fabric weave, architectural finish, food surface all hold up at 2K. 1K is purely a speed tradeoff.
Set aspect_ratio to 'custom' and pass width/height (256-2048 each) to get exactly the dimensions you need. Useful for hero banners, social spec sizes, and unusual ratios that don't fit the standard six presets.
Particularly strong on photographic genres — food photography, architectural exteriors, fashion editorial, product. Material rendering (fabric, concrete, metal, ceramic) reads clean and convincing without needing photo-stack technical anchors.
Each example shows the exact prompt that produced the result. Copy any prompt with one click.
2K · 16:9 · 1 credit
Wide architectural shot of a modernist concrete museum at golden hour, low sun raking across the angular geometric facade, deep cast shadows defining the form, single human figure for scale, hyper-detailed 8k photography
Architecture rewards 16:9 + golden hour + raking-light language. "Single human figure for scale" is a useful trick — Dreamina places the figure thoughtfully rather than dominating the frame.
2K · 1:1 · 1 credit
Overhead flat lay food photography of a Korean banchan spread on a dark slate surface — kimchi, japchae, banchan in small ceramic bowls, chopsticks resting on a linen napkin, dramatic side lighting, restaurant editorial quality
"Overhead flat lay" + named items + named surface + lighting direction is the canonical food editorial recipe. Dreamina renders ceramic, fabric, and food surface texture cleanly at 1:1 / 2K. Useful for menu shots and restaurant social.
2K · 3:4 · 1 credit
Fashion editorial portrait of a man with sharp features in a charcoal wool overcoat, urban concrete wall behind him, soft overcast lighting, looking directly at camera, magazine cover composition, photoreal
3:4 portrait is fashion-editorial sweet spot. Lead with subject + wardrobe + environment + lighting + composition. "Magazine cover composition" anchors framing without over-specifying the technical setup.
2K · 16:9 · 1 credit
Atmospheric fantasy scene of an ancient stone bridge crossing a misty ravine at dawn, two figures crossing in the distance, mountains rising beyond, painterly cinematic lighting, soft purple-blue palette with golden highlights
Dreamina excels at painterly cinematic atmospheres. Pair a clear focal element (the bridge) with foreground / midground / background depth and a defined color palette. 'Painterly cinematic' is a useful aesthetic anchor distinct from photoreal.
2K (4MP) costs the same as 1K. The only reason to drop to 1K is if you specifically need faster turnaround for prompt iteration. For any final output, generate at 2K.
"Golden hour", "overcast", "misty dawn", "painterly cinematic" — Dreamina reads atmospheric language as literal targets. These prompts hit harder here than on stricter photoreal models.
When you need exactly 1080×1350 (Instagram portrait) or 1920×1080 (banner spec), set aspect_ratio='custom' and pass width/height. Dimensions can be 256-2048 on either axis.
"Food photography", "architectural photography", "fashion editorial", "product editorial" — Dreamina has strong genre-specific composition behaviors. Naming the genre anchors framing, lighting, and depth-of-field conventions cleanly.
enhance_prompt=true rewrites your wording for richer detail. Useful when your prompt is short. Turn it off when you have precise wording you don't want paraphrased — particularly for in-image text.
Architectural and landscape shots benefit from "single human figure for scale" or "a person walking past" — Dreamina places the figure thoughtfully rather than dominating, and it sells the scale of the scene.
| Setting | Values | Notes |
|---|---|---|
| Resolution | 1K · 2K | Same credit cost. Default to 2K (4MP) for any final output. |
| Aspect ratio | 1:1 · 16:9 · 9:16 · 4:3 · 3:4 · custom | Six presets plus custom dimensions. |
| Custom dimensions | width / height 256-2048 each | Only used when aspect_ratio="custom". Useful for social spec sizes. |
| Enhance prompt | Boolean (default true) | Optional rewriting for richer detail. Off for literal prompt interpretation. |
| Seed | Integer (optional) | Pin for reproducibility. Omit for varied output across re-runs. |
| Pricing | Flat credit cost | Same across resolutions and aspects. |
Dreamina's defining trait is cinematic aesthetic defaults — output looks already-edited without needing photo-stack technical language. FLUX is sharper on hand rendering and prompt adherence. Seedream has more general-purpose aesthetic range. Qwen Image is stronger on text rendering. Pick Dreamina when you want film-still energy from sparse prompts.