ByteDance Seed · Video generation
Develop film scenes and advertising videos with a brief that directs action, camera and sound. Use Seedance 2.5 on Astorie to explore a narrative moment, a product reveal or a social creative, then review the clip before taking it into production.
Create video with native audio, reference-guided visuals and directed camera movement.
Give the subject a clear action and the camera a path to follow. Seedance 2.5 supports camera direction and performance blocking, making it a starting point for a moving shot rather than a single attractive frame. Describe the opening, turning point and final hold so the shot has somewhere to go.
Use references to guide a subject, setting or visual treatment instead of asking one image to control everything. Seedance 2.5 supports multimodal reference direction; Astorie’s documented Omni workflow connects visual references to the video node. State what each reference controls, then check the result for detail changes.
Seedance 2.5 generates audio and video together. Add dialogue, ambience or a specific sound cue alongside the visible action, and leave space for the moment to finish. Review the exported audio and picture together before approving speech, timing or synchronization.
Start creating with Seedance 2.5 in three simple steps.
Open an Astorie project and use the video toolbar to add a Video node. Prepare any subject or scene references, and connect the inputs needed for your shot.
Find Seedance 2.5 in the model menu. Check Omni mode, aspect ratio, resolution and duration in the current panel, then review model access and the displayed credit estimate.
Write the visible action, camera path and sound cues in the prompt editor. Generate when the settings are ready, then watch the completed clip with sound and inspect continuity before using it.
Develop product ads, previsualize film scenes and create tactile beauty videos for social.
Plan a reveal, a detail beat and a closing packshot around one visual idea. Keep the product readable as the motion changes, and leave a clean hold for approved campaign copy. The collage-style example shows one possible advertising treatment; inspect packaging and add final claims only after review.
Give the shot a change that matters: a character arrives, crosses a threshold and notices something. Use camera position, lighting and a final reaction to make the turn readable. Review the geography and performance before using the clip as a reference for the next production decision.
Focus a short beauty or lifestyle creative on one visible gesture. Specify the product texture, hand movement and final pose, then choose the supported ratio for the placement. Check skin, fingers and contact frames; create a separate version for each format rather than relying on an automatic crop.
Separate the model’s documented capabilities from the controls offered by Astorie. Check the selected node before committing a production brief.
| Provider and version | ByteDance Seed · Seedance 2.5.Source ↗ · 2026-11-06 |
|---|---|
| Main task | Generate moving scenes with native audio from a written brief and supported references.Source ↗ · 2026-11-06 |
| Reference direction | The provider documents image, video and audio reference support. Use the input types exposed by the selected Astorie workflow.Source ↗ · 2026-11-06 |
| Astorie model selection | The English canvas lists Seedance 2.5. The inspected Video node shows Omni mode.Source ↗ · 2026-11-06 |
| Output controls | The inspected Astorie node offers aspect ratio, resolution and duration controls. Its resolution menu lists 480p, 720p and 1080p; availability and estimates depend on the selected settings.Source ↗ · 2026-11-06 |
| Provider duration | Up to 30 seconds in one generation is documented by ByteDance. Use the duration offered in the current Astorie node; this page does not establish an extension workflow.Source ↗ · 2026-11-06 |
| Sound direction | Describe sound and dialogue in the brief. Astorie’s published Omni example records an audio track without a separate audio switch in the captured panel.Source ↗ · 2026-11-06 |
| Production review | Check identity, contact frames, product details, cuts, dialogue and sound in the actual export. Generated text and physical interactions can need correction.Source ↗ · 2026-11-06 |
Explore Seedance 2.5, Seedance 2.0 and Veo 3.1 through the inputs, creative direction and review steps each workflow needs.
| Dimension | Seedance 2.5 | Seedance 2.0 | Veo 3.1 |
|---|---|---|---|
| A direction to explore · 2026-11-06 | A scene with deliberate blocking and a clear narrative turn. | A scene guided by image, video or audio references. | An image-led shot with native sound. |
| How to shape the brief · 2026-11-06 | Define the opening, action, camera path and ending. | Assign each reference a subject, scene or motion role. | Describe the action, visual treatment and sound; use supported image inputs. |
| A shared review task · 2026-11-06 | Check the turn, continuity and exported audio. | Compare subject and scene details with the references. | Inspect image details, motion and sound together. |
| Before you generate · 2026-11-06 | Select Seedance 2.5 and inspect the current Omni controls. | Select the intended 2.0 variant; do not assume 2.5 settings. | Select Veo 3.1; check which provider features the node exposes. |
Use the clip to make a production decision, then review the details that matter to the final placement.
가장 강력한 AI 모델로 창작을 시작하세요. Free 요금제는 신용카드가 필요 없습니다.
3일 동안 Pro의 모든 기능 체험
Try another model when the next scene needs a different starting point. Inspect the specific variant and current canvas controls before running.
Seedance 2 is ByteDance's flagship multimodal video family, built on a unified audio-video joint generation architecture that accepts text, image, video and audio — up to 12 reference assets per generation. Six variants cover everything from a cost-effective base tier and original-price Offpeak mode to the full-featured Omni Pro, with six aspect ratios including ultra-wide 21:9.
Explore model
Google Veo 3.1 delivers industry-leading video generation with native audio synthesis, photorealistic quality, and resolution up to 1080p. The Extend variant adds video continuation and V2V capabilities for seamless clip extension.
Explore modelGo from a short beat sheet to a directed scene or a product ad. These guides cover wider workflows and may use other models.
To make short AI videos, define one hook and one payoff, prepare only the references each element needs, generate in the target aspect ratio, and audit the export before publishing. This tutorial shows the full process with a tested 15-second vertical UGC clip in Astorie.
Read guideLearn how to make a video cinematic by improving shot direction rather than relying on filters, black bars, or higher resolution. This before-and-after workflow shows how composition, lighting, motivated camera movement, and pacing transformed a simple AI-generated café scene.
Read guideLearn how to make a marketing video by defining your target audience, campaign goal, and distribution channel before production. This workflow shows how to turn a product brief into reusable visual assets, a timed storyboard, and a 9-second AI product ad using MiniMax H3.
Read guideModel choice, inputs, sound and review.
Seedance 2.5 is ByteDance Seed’s audio-video generation model. On Astorie, choose it in a Video node to explore a moving scene from a written brief and supported references.
Generate in the Astorie canvas. This page explains the model and offers creative starting points; it does not contain a video generator.
No. Open the canvas, select Seedance 2.5 and copy the brief into the prompt field. The button opens Astorie; it does not configure a node or transfer the text.
Give each connected reference one role, such as character, product, room or opening composition. State that role in the brief and compare the generated result with the source. Astorie’s linked short-video guide documents a four-reference Omni workflow; it is not a universal input limit.
Write the intended words, ambience and sound cues alongside the action. Check the current node for any additional audio controls, then listen to the exported clip. A prompt is not proof that the words or synchronization are correct.
ByteDance documents up to 30 seconds in a single generation. Choose a duration available in your current Astorie node, and keep the action simple enough to finish within it.
Choose a supported aspect ratio and rewrite the framing for each placement. Generate and review each version separately, checking that faces, products and later copy space stay within the intended composition.
Read the estimate in the selected node before running. The shared subscription section describes Astorie plans; it is not a fixed Seedance 2.5 generation price. Access and estimates can change with model and output settings.
Adapt each prompt to your subject, action and format, then review the generated clip. The videos are existing Astorie library examples; the prompt starters offer new directions rather than their original generation prompts.
Watch the full export with sound, then inspect the image sequence for continuity, contact frames, text and product detail. Verify spoken lines, usage permissions and delivery requirements before the clip becomes a final production asset.
Set the action, direct the camera and sound, then review the clip before it joins the production.
