PRACTICAL GUIDE · PROMPTTOVISUAL
Seedream 5.0 Flash: API Pricing, References and Editing Guide
Seedream 5.0 Flash is ByteDance’s faster, lower-cost image generation and editing tier. This independent guide compares provider prices and input limits. PromptToVisual has not integrated or tested Flash. Sources checked October 5, 2026.
Official BytePlus tutorial · Runware API · OpenRouter listing · PTV editor: different models · All guides
Where is Flash available?
BytePlus lists dola-seedream-5-0-flash-260915 in its September 2026 release table. The suffix is an identifier, not proof of a September 15 launch. Runware uses bytedance:seedream@5.0-flash; OpenRouter uses bytedance-seed/seedream-5-0-flash. Confirm access in your chosen account before planning a paid workflow.
Flash versus Pro
BytePlus positions Flash for lower latency and cost, and Pro for image quality. Both support generation, reference editing, location-based edits and layer decomposition. Its tutorial lists 1K, 1.5K and 2K outputs for both; Pro offers standard and fast prompt optimization while Flash lists standard. This is vendor positioning, not our measured comparison.
BytePlus API prices: generation and layers differ
The official table lists Flash at $0.018 per output image with free image input. Pro single-image output costs $0.045 up to 2.61 million pixels and $0.09 above that; its first input is free and subsequent inputs cost $0.003 each. Pro layers cost $0.0225 or $0.045 each by that pixel boundary. Do not apply a single-image rate to an entire multi-layer request or assume Pro layer discounts apply to Flash. Confirm Flash layer billing before estimating a multi-output job.
Third-party prices are not PromptToVisual prices
Runware lists $0.01926 per output, displayed as $0.0193 per image. Its example costs do not guarantee a total for every layer request. OpenRouter lists $0.018 per image. These are provider API offers; PTV has no Flash generation offer. Prices checked October 5, 2026; account terms and requested operation still matter.
Reference limits depend on the provider
Runware allows up to 10 references for ordinary editing; transparent-background and decomposition modes each allow one. OpenRouter’s listing says up to 14 references. That discrepancy is not a universal guarantee: verify the exact provider contract instead of moving a payload between services unchanged.
Resolution and request mode
Runware documents width/height pairs within 0.92–4.62 megapixels and aspect ratios from 1:16 to 16:1. Supply both dimensions. Its resolution preset belongs to decomposition, where width/height and positivePrompt are excluded. Ordinary generation requires positivePrompt. BytePlus uses a different request shape; do not transplant these fields into its API.
An original editing prompt — untested
For a ceramic mug photo: “Change only the mug glaze to matte teal. Keep the handle shape, rim, position, lighting and table unchanged. Do not add text or objects.” This original example has not been run. Inspect the whole result: preservation instructions are not pixel locks. Use a conventional editor for exact geometry or lettering.
Make a local edit easy to inspect
Name the object, one requested change and the details to retain. Avoid combining recoloring, relighting and a new background in the first attempt. BytePlus documents coordinates and drawn markers; follow its current coordinate conventions. Check edges, reflections, spelling and whether any markers remain.
Layer decomposition is a separate workflow
BytePlus describes one base image plus up to 16 alpha layers, with position and stacking metadata. That is not proof of a finished editable design project or perfect separation of overlapping objects. Inspect edges and reconstruction. Flat-image generation and decomposition need separate requests and cost checks.
Choose a next step
Use the external provider links for Flash itself. PromptToVisual’s image editor and model directory offer other models with their own credit costs. They are alternatives for an image task, not Flash access. No Flash sample, matched benchmark or integration was performed for this guide.
Continue with the right tool
AI image combiner · Stitch images · Photo collage · Nano Banana 2 guide · Related composition guide