文字或圖像
從鏡頭簡介開始，或使用經過批准的視覺錨定主題和構圖。
Native 30-second multimodal production
← Seedance 模型中心
Seedance 2.5 人工智慧視訊產生器
Generate continuous single-shot video up to 30 seconds in one pass, with up to 50 multimodal references, native 4K, and region-level editing — no stitching, no sequence breaks.
Long-form multimodal routeUp to 30 seconds in a single shot480p, 720p, 1080p, native 4K
免費開始創作精確路線工作空間
生成為 Seedance 2.5
目前的控制和收費積分來自即時產生器。提交前確認它們。
載入中...
Long-form multimodal route
Longer shots, more references, surgical edits
Seedance 2.5 extends the 2.0 multimodal architecture: longer native duration, the highest reference limit among commercial models, and region-level editing that preserves overall consistency.
01
30-second single-shot films
Direct a continuous 30-second take in one generation instead of stitching multiple 5–15s clips.
用它來回答： 這個方向是否足夠清晰，可以繼續下去，哪些細節必須在下一輪中保持固定？
02
50 multimodal references
Combine up to 50 image, video, text, and audio inputs to control characters, scenes, style, and sound together.
用它來回答： 這個方向是否足夠清晰，可以繼續下去，哪些細節必須在下一輪中保持固定？
03
Region-level editing
Change a product, prop, or background area while the rest of the clip keeps its motion, lighting, and identity.
用它來回答： 這個方向是否足夠清晰，可以繼續下去，哪些細節必須在下一輪中保持固定？
04
Native 4K delivery
Output crisp 4K natively for large displays, product films, and high-DPI social.
用它來回答： 這個方向是否足夠清晰，可以繼續下去，哪些細節必須在下一輪中保持固定？
05
Character dialogue across long takes
Keep identity, eyelines, and lip-sync stable while directing a 30-second performance with linked room sound.
用它來回答： 這個方向是否足夠清晰，可以繼續下去，哪些細節必須在下一輪中保持固定？
06
Social hooks for vertical delivery
Test opening beats as controlled variants while preserving the same product, palette, and final packshot.
用它來回答： 這個方向是否足夠清晰，可以繼續下去，哪些細節必須在下一輪中保持固定？
路線比較
按製作角色選擇 Seedance
該表將目前的 PixMind 路線快照與位元組跳動家族層級的聲明分開。應在生成器中重新檢查積分和公開的控制項。
| 版本 | 最適合 | 輸入 | 持續時間 | 解析度 | 音訊 | 信用快照 |
|---|---|---|---|---|---|---|
| Seedance 2.0 Pro | 複雜的多模式參考和選定的最終鏡頭 | 文字、圖像、視訊和音訊參考 | 5、10 或 15 秒 | 480p、720p、1080p； 4K需要即時驗證 | 支援音訊的路線 | 從實時生成器中讀取 |
| Seedance 2.5目前 | Native 30-second single-shot video with up to 50 multimodal references | Text, image, video, and audio references (up to 50) | Up to 30 seconds in a single shot | 480p, 720p, 1080p, native 4K | 支援音訊的路線 | 從實時生成器中讀取 |
| Seedance 2.0 Fast | 分鏡節拍、攝影機 A/B 測試和社交掛鉤 | 文字、圖像、視訊和音訊參考 | 5、10 或 15 秒 | 480p 或 720p | 支援音訊的路線 | 從實時生成器中讀取 |
| Seedance 2.0 Mini | 測試參考擬合、組成和提示衝突 | 文字、圖像、視訊和音訊參考 | 5、10 或 15 秒 | 480p 或 720p | 支援音訊的路線 | 從實時生成器中讀取 |
| Seedance 1.5 Pro | 現有的影像引導提示和受控遷移測試 | 當前回應中的文字或最多兩張圖像 | 現場生成器是最終的； API字段衝突 | 480p、720p 或 1080p | 支援音訊的路線 | 40點API快照；確認直播 |
1.5 Pro 回應報告 40 分，但其持續時間欄位有衝突。 Pro、Fast 和 Mini 不會在已驗證的端點快照中公開穩定點欄位。
測試後要注意什麼
將幾代人轉變為生產決策
來源影像設定了天花板
使用完整的主題、乾淨的邊緣、可讀的深度以及足夠的空間來滿足所要求的移動。
受限運動保護身分和幾何形狀
將主體的動作與攝影機的移動分開，避免在一張短鏡頭中疊加多個較大的變化。
出於生產需求而不是版本號進行遷移
當鏡頭需要多模式參考、第一幀/最後一幀或更明確的音訊方向時，請移至 2.0 Pro。
如何使用該版本
從原始資料到審閱的內容
Assign one job per reference
Give every image, video, text, or audio input a single explicit role (identity, shape, motion, palette, or rhythm).
Lock the continuity anchors
Fix identity, wardrobe, props, screen direction, and the ending composition before adding complex motion.
Use region edits sparingly
Edit one region per pass and review identity, geometry, lighting, and timing in the full clip each time.
Confirm live pricing
Read the live generator for the current credits at the chosen duration and resolution before submitting.
特定於版本的提示
提示此路線的啟動者
提示 01
30-second product film
Use the full native duration for one continuous reveal.
One continuous 30-second take. Use @Image 1 for exact product shape and @Image 2 for the studio palette. Slow controlled orbit around the product, one subtle environmental interaction at the midpoint, finish on the approved hero composition. Preserve shape, material, ports, and color. No generated text.
提示 02
Multimodal character scene
Coordinate identity, motion, and sound across references.
Keep the character from @Image 1 unchanged. Use @Video 1 only for the body motion and @Audio 1 for the ambient rhythm. The character walks through the station, pauses, and looks toward the exit. Preserve facial features, wardrobe, and left-to-right screen direction.
提示 03
Region edit pass
Change one area without re-generating the shot.
Keep the full clip motion, lighting, and identity fixed. Only replace the background product on the right shelf with the object in @Image 2, matching perspective, scale, and reflections. Do not alter the subject or camera path.
比較已連結的 Seedance 版本
Pro
Seedance 2.0 Pro
Quality-first multimodal video with reference and audio control.
查看版本Pro
Seedance 2.5
Native 30-second single-shot video with up to 50 multimodal references, 4K output, and region-level editing.
Fast
Seedance 2.0 Fast
更快的 Seedance 2.0 路線，適用於迭代繁重的社交、行銷活動和預視覺化工作流程。
查看版本Mini
Seedance 2.0 Mini
一個更輕的 Seedance 2.0 選項，用於在致力於品質第一渲染之前測試多模式想法。
查看版本Pro
Seedance 1.5 Pro
一種經過驗證的影像到視訊選項，具有廣泛的縱橫比和解析度控制，可實現直接的參考主導鏡頭。
查看版本Seedance 2.5 FAQ
What is new in Seedance 2.5?
Native single-shot video up to 30 seconds, up to 50 multimodal references (image, video, text, audio), region-level editing that preserves consistency, and native 4K output.
How is Seedance 2.5 different from 2.0?
2.0 generates 5–15s clips with up to 9 references; 2.5 extends single-shot duration to 30 seconds, raises the reference limit to 50, and adds region-level editing.
How do the 50 multimodal references work?
Pass up to 50 inputs across image, video, text, and audio in one request. Give each a single explicit role and remove assets that compete for the same property.
Can I edit part of a video without regenerating it?
Yes. Region-level editing lets you change a specific area while the rest of the clip keeps its motion, lighting, and identity. Edit one region per pass and review the full clip each time.
What duration and resolution does Seedance 2.5 support?
Single-shot clips up to 30 seconds and resolutions up to native 4K. Confirm the exact options in the live generator before submitting.
How much does Seedance 2.5 cost?
Pricing is not yet published. Read the live generator for the current credits at your chosen duration and resolution before submitting.
Does Seedance 2.5 generate audio together with the video?
Yes. The unified audio-video architecture generates synchronized audio in the same task when the selected mode supports it. Review speech, lip movement, source matching, distortion, and timing before publishing.
What should I review before publishing a Seedance 2.5 clip?
Review identity, faces, hands, product geometry, text, logos, continuity, dialogue, and audio timing frame by frame across the full clip. Confirm rights for every uploaded person, brand, voice, image, video, and audio asset.
下一個生產步驟
比較每條 Seedance 路線
Return to the family hub for route selection, current parameter notes, and migration guidance.