
Seedream v5.0 Flash Text-to-Image API by ByteDance
Dola-Seedream-5.0-flash, ByteDance's high-speed, cost-optimized image generation model built for instant creation. Generates a single image from a text prompt in seconds at 1K, 1.5K, or 2K - well suited to latency-sensitive scenarios such as mobile assistant apps, UGC content creation, and bulk-scale content production.
Seedream v5.0 Flash Text-to-Image 由 ByteDance 開發。Atlas Cloud(由 Atlas Cloud AI LLC 營運)僅提供接取服務,並不擁有該模型。所有商標均歸其各自所有者所有。
1. Introduction
Seedream 5.0 Flash (Dola-Seedream-5.0-flash) is ByteDance's high-speed, cost-optimized member of the Seedream 5.0 image generation family, added in September 2026 (upstream model version dola-seedream-5-0-flash-260915). This README covers the API model identifier:
bytedance/seedream-v5.0-flash/text-to-image
Seedream 5.0 Flash shares Seedream 5.0 Pro's request contract, resolution envelope, and multilingual prompt support, but is tuned for latency and volume: a single 2K image typically completes in 10–20 seconds. ByteDance positions Dola-Seedream-5.0-flash for instant creation: a high-speed, cost-optimized model covering text-to-image generation, single- and multi-reference image workflows, precision editing, and layer decomposition, aimed at latency-sensitive scenarios such as mobile assistant applications, UGC content creation, and bulk-scale content production. (BytePlus ModelArk)
2. Key Features & Innovations
-
Fast Single-Image Generation: Returns one image per request, typically in 10–20 seconds at 2K and faster at 1K/1.5K, without a slow deep-thinking pass — well suited to interactive tools and on-demand pipelines.
-
Pro-Class Resolution Envelope: Output from 921,600 (1280×720) to 4,624,220 total pixels with any aspect ratio from 1:16 to 16:1, including official 2K presets such as 2048×2048, 2816×1584 (16:9), and 3136×1344 (21:9).
-
Multilingual Prompts: Beyond Chinese and English, prompts are understood in Russian, Arabic, Filipino, Thai, Turkish, Korean, Malay, Spanish, Portuguese, Indonesian, French, German, Vietnamese, and Japanese.
-
Clean Typography: Inherits the Seedream 5.0 family's text-rendering strength for signage, posters, and labels rendered directly into the image.
-
PNG and JPEG Output: Choose lossless PNG for design handoff or compact JPEG for web delivery; results are returned as hosted URLs or Base64.
3. Model Architecture & Technical Details
Seedream 5.0 Flash is built on the Seedream 5.0 family's multimodal transformer architecture — vision encoders coupled with a diffusion-based image decoder — with an inference path optimized for speed. Automatic prompt optimization runs in standard mode; the family's fast optimization mode and the ability to switch off prompt thinking are Pro-only controls, since Flash is already tuned for low latency.
Size is selected either by explicit WIDTH*HEIGHT (total pixels within [921,600, 4,624,220], aspect ratio within [1/16, 16]) or by the tier keywords 1K, 1.5K, and 2K, in which case the model infers the aspect ratio from the prompt. Sequential (multi-image) generation and streaming output are not supported: every request produces exactly one image.
4. Performance Highlights
Public arena rankings for the Flash snapshot are not yet available; its position in the Seedream 5.0 family:
| Model | Developer | Output Range | Reference Images | Typical Latency (2K) |
|---|---|---|---|---|
| Seedream 5.0 Flash | ByteDance | 1K–2K | up to 10 | 10–20 s |
| Seedream 5.0 Pro | ByteDance | 1K–2K | up to 10 | slower with prompt thinking enabled |
| Seedream 5.0 Lite | ByteDance | 2K–4K | up to 14 | — |
Flash keeps Pro's single-image discipline and output envelope while cutting generation time, making it the default pick when throughput and responsiveness matter more than squeezing the last increment of per-image quality.
5. Intended Use & Applications
-
Mobile Assistant Applications: Instant image generation inside chat and assistant apps, where every second of latency is felt by the user.
-
UGC Content Creation: Fast, on-demand visuals for user-generated content platforms — avatars, posts, stickers, and creative prompts at interactive speed.
-
Bulk-Scale Content Production: Social posts, thumbnails, ad variations, and catalog imagery generated at scale with fast turnaround at any resolution.
-
Rapid Concept Exploration: Fast iteration on composition, style, and copy before committing a final render.
-
Localized Marketing Visuals: Multilingual prompts for regional campaigns, with text rendered directly into posters and banners.
-
E-Commerce Imagery: Product scenes and lifestyle shots at 2K for listing pages and advertising.



















