Seedream 5.0 Flash: Pricing, Multi-Reference Editing & API Guide

Seedream 5.0 Flash is ByteDance’s model in the Seedream 5.0 family for efficient image generation and editing. On Poyo, both workflows cost 4 credits ($0.02) per image, with 1K, 1.5K, and 2K output and up to 10 reference images for editing. The official release notes position it for workloads sensitive to latency and cost.

To explore different versions of a product, character, or multilingual asset, generate candidates first, then refine your selection with references. Through ingredient illustrations, portrait scenes, and a porcelain display, this guide explains reference roles, prompt writing, creative budgets, and the playground and API steps.

Seedream 5.0 Flash Pricing, Output, and Editing Settings at a Glance

Item	Feature on Poyo
Provider	ByteDance
Text-to-image model	seedream-5.0-flash
Image editing model	seedream-5.0-flash-edit
Output resolution	1K, 1.5K, 2K; API default 1K
Aspect ratios	1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9
Editing references	1–10 images
Output per request	One image
Optional output formats	JPEG, PNG
Safety checker parameter	enable_safety_checker
Generation and editing price	4 credits / $0.02 per image

Try generation and editing on the Seedream 5.0 Flash model page. The API documentation covers the complete parameters.

Which Workflows Suit Seedream 5.0 Flash?

Repeated Creative Exploration

Keep the subject and style fixed while adjusting backgrounds, colors, framing, or copy across rounds to build a set of comparable candidates. Flash’s cost and efficiency focus suits this approach to drafting. Teams can turn each round’s brief into a prompt and record its output and price.

Multilingual Visual Content

Seedream 5.0 Flash supports multilingual generation. For illustrated cards, exhibit labels, and localized visuals, specify the language, exact copy, and layout hierarchy together. Clear positions for the title, illustration, and supporting text help express the creative brief.

Multi-Reference Editing

The edit model can combine 1–10 references. Supply a person, product, and environment separately, then describe their relationship for portrait scenes, product background changes, and object composition. Specify the subject’s silhouette, materials, clothing, and scene structure, along with lighting and perspective requirements.

Seedream 5.0 Flash vs. Lite vs. Pro

Flash, Lite, and Pro are models within ByteDance’s Seedream 5.0 family. Choose by iteration budget, output dimensions, and the complexity of the creative brief.

Start with Flash when exploring several 1K–2K candidates and planning a clear per-round budget. Choose Lite when delivery calls for 3K or specific dimensions. For a dense layout or precise editing brief, test Pro with the same references and copy, then assess composition, lettering, and subject integration against your requirements.

Seedream 5.0 Flash Examples: Multilingual Illustration and Multi-Reference Editing

These inputs and outputs come from the generation examples on Poyo’s model page. Each example explains the reference roles, requested aspect ratio and resolution, and details to inspect in the result.

1. Ingredient Notes in Three Languages

Input Prompt excerpt Create a beautifully illustrated kitchen ingredient notebook page, landscape 4:3, warm ivory paper, delicate watercolor with crisp readable typesetting. Two clearly separated columns: left contains one ripe tomato, one sliced tomato and green leaves; right contains a small bunch of fresh basil leaves. At the top write exactly '食材笔记' with 'Ingredient Notes' below it. Beneath the tomato illustration print three centered lines exactly: '番茄', 'Tomato', 'トマト'. Beneath the basil illustration print three centered lines exactly: '罗勒', 'Basil', 'バジル'. Chinese, English and Japanese must be legible and correctly spelled, each label associated only with its own illustration. Restrained serif English, clean Chinese and Japanese lettering, balanced margins and ample spacing.

Output

Input and settings: Generate from a text prompt at 4:3 and 2K. The prompt places tomatoes and basil in separate columns and specifies warm ivory paper, watercolor illustrations, and labels line by line.

What to inspect: The title reads “食材笔记 / Ingredient Notes”, with Chinese, English, and Japanese labels underneath. Check the relationship between each ingredient and its labels, then review title hierarchy, letter spacing, and margins.

Prompt approach: Define the two-column composition, supply each label, and then describe typography, spacing, and margins. Use the same element-to-label structure for recipe cards, learning materials, and multilingual product introductions. Proofread each line and check its relationship to the illustration before delivery.

2. Place a Person in an Alpine Station

Input Prompt excerpt Create a photorealistic travel documentary photograph using both references. Place the exact woman from reference 1 on the clear foreground platform in reference 2. Preserve her facial identity, short dark wavy bob, mustard-yellow jacket, burgundy knitted scarf, charcoal trousers, brown boots and steel-gray backpack. Show her full body standing naturally slightly left of center, face visible, both boots resting on the paving. Keep the old stone station to the left, green train to the right and snowy alpine mountains in the background. Match the overcast cool environmental light, camera perspective, subject scale and soft contact shadows on the wet stone.

Output

Input and settings: Image 1 supplies the character and image 2 the station. Generate one travel photo at 3:4 and 1.5K.

What to inspect: Compare the face, yellow jacket, and burgundy scarf with image 1, then the stone station, green train, and snowy mountains with image 2. Finally, check the scale, lighting, and contact shadows between the person and platform.

Prompt approach: Assign separate sources for the person and environment, then specify placement, full-body framing, light, scale, and contact shadows. This division of reference roles suits travel concepts, portrait scenes, and clothing visuals.

3. Place Porcelain in a Display Case

Input Prompt excerpt Combine the two references into one photorealistic museum exhibit photograph. Take the blue-and-white porcelain vase from reference 1 and place this exact vase upright at the center of the empty glass vitrine in reference 2. Preserve its silhouette, cobalt peony patterns, narrow neck, geometric rim band and glossy white glaze. Preserve the museum architecture and oak vitrine. Match the warm gallery lighting, realistic reflections through glass, scale, perspective and a soft contact shadow on the pedestal. Replace the blank label with a clean cream exhibit card reading exactly two centered lines: '青花瓷' and 'Blue-and-white Porcelain'. Use elegant dark museum lettering, clearly legible, with the Chinese above the English. Show the complete vase and vitrine in a refined 4:3 documentary composition.

Output

Input and settings: Image 1 supplies the vase and image 2 the empty display case. Select 4:3 and 2K.

What to inspect: The output places the porcelain vase in the center of the glass case, with a wooden cabinet, warm gallery lighting, and the label “青花瓷 / Blue-and-white Porcelain”. Compare the vase silhouette and cobalt patterns with the reference, then inspect its scale, position, and reflections inside the case.

Prompt approach: Specify the vase silhouette, cobalt patterns, glaze, and cabinet structure to preserve, then define placement, proportions, and reflections. Adapt this structure to your own objects and environments for product displays and exhibition concepts.

Seedream 5.0 Flash Multi-Reference Editing: Assign a Role to Each Image

Step 1: Prepare and Order Your References

Start with the subject and target scene, then add detail or style references as your brief requires. Poyo editing supports 1–10 references, and the prompt can assign each image a specific role:

Reference role	What it supplies	How to describe it
Subject	Character face, product silhouette, or object patterns	“Image 1 supplies the character and clothing” or “Image 1 supplies the vase”
Scene	Background structure, viewpoint, placement, and lighting	“Image 2 supplies the station” or “Image 2 supplies the display case”
Detail	Materials, textures, markings, or clothing details	Identify the specific details to preserve from additional images
Style	Color palette, photographic style, or illustration strokes	Identify the style reference and the visual traits to adopt

Keep the upload order aligned with the image numbers in your prompt. In the API, the first entry in image_urls is image 1 and the second is image 2. You can retain the same order in later edits as well.

Step 2: Describe the Reference Relationships and Final Scene

Use “reference roles + final scene + features to preserve + environment matching + text layout.” For the display-case example, adapt this prompt template to your own object and scene:

Place the blue-and-white porcelain vase from image 1 at the center of the glass display case in image 2. Preserve the vase silhouette, cobalt patterns, and glaze, along with the museum architecture and wooden cabinet. Match the vase’s scale, perspective, warm lighting, and contact shadow to the case. Add a label in front of the vase with two centered lines: “青花瓷” and “Blue-and-white Porcelain”.

For portrait scenes, use the same structure: identify the character and background sources, then specify placement, pose, clothing, framing, and lighting. For text tasks, provide copy line by line and describe the subject and position associated with each text block.

Step 3: Refine One Objective in the Next Round

After choosing a candidate, use it as the next round’s reference and adjust subject placement, background, or text layout in turn. Describe this round’s changes and the features to preserve, then inspect the face, product patterns, perspective, and shadows. Record the prompt and settings with each image to compare versions.

Seedream 5.0 Flash Pricing and Creative Budgets

These Poyo prices were updated on October 8, 2026. All three resolution tiers share the same price per image:

Workflow	Model ID	Resolution	Price per image
Text to image	seedream-5.0-flash	1K, 1.5K, 2K	4 credits / $0.02
Image editing	seedream-5.0-flash-edit	1K, 1.5K, 2K	4 credits / $0.02

By generated image count, 100 images cost 400 credits ($2) and 1,000 images cost 4,000 credits ($20). For example, generate five candidates for each of 20 concepts, then edit one selected image per concept once. That produces 120 images with a budget of 480 credits ($2.40).

Each request produces one image. Developers can organize multiple requests into a batch workflow and track each output by task ID. Choose resolution by delivery size: 1K for previews, 1.5K for medium-sized assets, and 2K for higher-resolution delivery. See the model page for current pricing.

How to Use Seedream 5.0 Flash in the Poyo Playground

1. Open the Seedream 5.0 Flash model page.
2. Select text-to-image generation or image editing.
3. Describe the subject, composition, style, lighting, and exact copy.
4. For editing, upload 1–10 references and explain each image’s role.
5. Set the aspect ratio, 1K / 1.5K / 2K resolution, and desired output format.
6. Generate, then inspect the text, object details, and character features, then refine the prompt and download your selection.

Keep the subject fixed and adjust the background, colors, and framing in sequence to create a clear record of your iterations. Compare the visual results, how closely each image meets the brief, and the cost per round when choosing the final direction.

Seedream 5.0 Flash API Integration

The following examples demonstrate the request structure, using simplified prompts based on the ingredient illustration and display-case editing tasks.

Submit a Text-to-Image Request

Create a Poyo API key, then call the unified submission endpoint:

curl --request POST \
  --url https://api.poyo.ai/api/generate/submit \
  --header 'Authorization: Bearer YOUR_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "seedream-5.0-flash",
    "input": {
      "prompt": "Create a watercolor ingredient card on warm ivory paper. Use two columns: tomato on the left and basil on the right. At the top write 食材笔记 and Ingredient Notes. Under the tomato write 番茄, Tomato, トマト. Under the basil write 罗勒, Basil, バジル. Use crisp readable lettering, generous spacing and balanced margins.",
      "size": "4:3",
      "resolution": "2K",
      "output_format": "png",
      "enable_safety_checker": true
    }
  }'

After a successful submission, read the task ID from data.task_id in the response.

Submit a Multi-Reference Edit

Use seedream-5.0-flash-edit for editing, with references supplied in order in image_urls. This request body places the vase into the display case:

{
  "model": "seedream-5.0-flash-edit",
  "input": {
    "prompt": "Place the blue-and-white porcelain vase from image 1 at the center of the glass vitrine in image 2. Preserve the vase silhouette, cobalt floral patterns and glossy glaze, and preserve the museum architecture. Match the warm lighting, perspective, scale and contact shadow. Add an exhibit card with two centered lines: 青花瓷 and Blue-and-white Porcelain.",
    "image_urls": [
      "https://storage.poyo.ai/models/seedream-5-0-flash/examples/museum-input-1.webp",
      "https://storage.poyo.ai/models/seedream-5-0-flash/examples/museum-input-2.webp"
    ],
    "size": "4:3",
    "resolution": "2K",
    "output_format": "png",
    "enable_safety_checker": true
  }
}

size sets aspect ratio and defaults to 1:1. resolution sets resolution and defaults to 1K. Choose jpeg or png with output_format and configure safety checking with enable_safety_checker. Prompts accept 1–5000 characters.

Poll for Results or Receive a Callback

Query status with the task ID returned by the submission:

curl --request GET \
  --url https://api.poyo.ai/api/generate/status/YOUR_TASK_ID \
  --header 'Authorization: Bearer YOUR_API_KEY'

Poll every 2–5 seconds. When data.status becomes finished, read each image’s file_url from data.files and save it. You can also add a top-level callback_url to the submission request to receive a notification when the task completes. The task status documentation covers the full flow.

Frequently Asked Questions

What is Seedream 5.0 Flash?

Seedream 5.0 Flash is ByteDance’s cost-efficient image generation and editing model in the Seedream 5.0 family. It suits concept iteration, multilingual visuals, and multi-reference editing. On Poyo, it offers 1K, 1.5K, and 2K output at 4 credits ($0.02) per image.

Which tasks suit Flash, Lite, and Pro?

Use Flash for repeated concept exploration and multi-reference edits at 1K–2K. Lite supports 2K, 3K, and custom dimensions for briefs with specific output sizes. Pro focuses on complex prompts, dense layouts, multilingual typography, and precise regional edits. Compare candidates using the same copy and references when choosing a model for your brief.

How much do generation and editing cost on Poyo?

Text-to-image generation and reference-based editing both cost 4 credits ($0.02) per image, across 1K, 1.5K, and 2K. Reference editing accepts up to 10 images at that per-output price. For example, 100 generated images cost 400 credits ($2); include candidate images and later edits in the total.

How should I choose 1K, 1.5K, or 2K, and how does the price differ?

All three resolutions cost 4 credits ($0.02) per image. Use 1K for previews and composition exploration, 1.5K for medium-sized assets, and 2K for larger display sizes or detail review. Choose by the final placement and viewing size; the API defaults to 1K resolution.

How many references can I use, and how should I order them?

Poyo’s playground and editing API support up to 10 reference images. Upload them in the order used in your prompt and assign a role to each: for example, image 1 supplies the person or product and image 2 supplies the scene. Describe the features to preserve, the subject’s placement, and the intended lighting.

How can I place a person or product naturally into a new scene?

Provide a clear subject reference and a scene reference, then identify their roles in the prompt. Specify placement, scale, camera angle, light direction, contact shadows, and relevant reflections. The railway and museum examples above show this structure; inspect the subject’s features and its relationship to the surroundings before refining the next round.

How should I write prompts for multilingual text and complex scenes?

Begin with the subject, scene, composition, and lighting, then describe each element and its location. Put exact in-image text in quotation marks and assign its language, position, hierarchy, and font style. For multiple references, state what each contributes. After generation, check the text line by line and review the relationships between objects.

How can I plan repeated drafts and batch creative work with Flash?

Group variations by concept and keep shared subject, style, and reference instructions consistent. Change one creative variable per round, such as framing, background, or lighting, and save the prompt alongside the chosen result. Budget by total outputs: three candidates for each of 10 concepts produce 30 images and cost 120 credits ($0.60); further edits add to that total. The API workflow above covers individual submissions and result tracking.

How can I try Seedream 5.0 Flash on Poyo and download the result?

Open the Seedream 5.0 Flash model page, choose generation or editing, and enter your prompt. For editing, upload references in the stated order. Select the aspect ratio and resolution, generate, then inspect and download the completed image; use it as a reference for a further editing round.

Start on the Seedream 5.0 Flash model page, choose an example, replace the subject or references, and refine your visual concept across rounds.