Seedance 2.5

Seedance 2.5

Seedance 2.5 is ByteDance's next-generation video generation model. It generates one native 30-second single-shot clip per request, with audio generated in the same pass, and accepts up to 50 multimodal references - 30 images, 10 videos, and 10 audio clips. References can anchor subjects, motion, camera behavior, style, and sound, while timestamp-level instructions direct how they combine across the clip.

Text to Video, Image to Video, Reference to Video, and First & Last Frame tasks are all available on MachGen, with white-model and motion references for staging and camera control.

Why choose Seedance 2.5

Key strengths

• One native 30-second single-shot clip per request - a complete arc without stitching separate clips.
• Up to 50 multimodal references per request - 30 images, 10 videos, and 10 audio clips - so complex scenes can anchor many subjects, sounds, and styles at once.
• Reference types go beyond stills and clips: white-model (3D blockout), motion, and creative references can steer composition, camera, and visual intent.
• Audio-native output: dialogue, ambience, and sound effects are generated in the same pass and stay synchronized with the footage.

Good fit for

• Long narrative scenes that need multiple logically connected shots in one pass.
• Multi-subject scenes - groups, ensembles, products, and locations - where many references must stay consistent together.
• Pre-visualization from white-model layouts before committing to full renders.
• Audio-first scenes where synchronized dialogue, ambience, and sound design matter as much as the visuals.

Inputs and output settings

This table reflects the model routes and controls currently exposed by MachGen.

Using Seedance 2.5

1. Choose a supported generation mode in Playground.
2. Add the required prompt and source or reference assets.
3. Select output settings from the controls shown for that mode.
4. Generate, then inspect the returned asset before reusing the settings through the API.

Practical guidance

• Describe the subject, action or composition, environment, and visual direction in a clear order.
• For image animation, describe the intended motion and camera behavior instead of repeating what is already visible.
• Give each reference a clear role in the prompt, using the names inserted by the Playground.
• Use compatible start and end frames, then describe the transition between them.
• When audio is enabled, include any dialogue, ambience, or sound cues that matter to the scene.

Up to 50 references in one request

Seedance 2.5 accepts up to 30 images, 10 videos, and 10 audio clips per request - 50 total. The reference board in MachGen's Generate surface replaces the one-row-per-file layout with a compact grid, bulk upload per kind, and a tile editor for naming, replacing, and removing assets.

• Images anchor characters, products, locations, styles, and materials.
• Videos contribute motion, camera behavior, scene continuity, and editing rhythm.
• Audio anchors voice, music, ambience, and sound design.
• White-model and motion references steer staging, camera path, and subject trajectory.

Editing and extending a clip

Editing and extending an existing video run through this same surface. Attach the clip as a video reference and set framing to Adaptive - the model reads your prompt to decide whether you are editing the clip or continuing it, and Adaptive is the only framing either accepts. Leave the duration on Match source to edit, where the result runs exactly as long as the clip you attached; pick a length instead to extend past it. With several videos attached, the first is the one being edited and the rest only inform the result.

Prompting notes

• Give every reference an explicit role with @name, and say which one controls subject, scene, motion, style, voice, or rhythm.
• For long clips, write beats in chronological order with timestamps so story, camera, and audio map to the right seconds.
• Describe one main action and camera move per shot; Seedance 2.5 can organize several logically connected shots inside one 30-second clip.
• Keep stable visual anchors - clothing, silhouette, lighting, material - across references so multi-subject scenes stay consistent.

Input and output limits

The official envelope: 30 images, 10 videos, and 10 audio clips, 50 references combined, and 4-30 second output at 24 fps. Output resolution is 480p or 720p (official 2.5 does not support 1080p or 4K); aspect ratios are 16:9, 9:16, 1:1, 4:3, 3:4, and 21:9, plus Adaptive, which keeps the reference material's own framing. First-frame and first-and-last-frame tasks are fixed to Adaptive, and reference-to-video offers it alongside the fixed ratios.

Per-file limits follow the official API: images below 30 MB at 300-6000 px per axis with an aspect ratio from 2:5 to 5:2; videos as MP4 or MOV, 2-30 seconds each with 30 seconds combined, 24-60 fps, 409,600-8,295,044 total pixels, and below 200 MB; audio as WAV or MP3, 2-30 seconds each with 30 seconds combined, and below 15 MB. The MachGen request shape is the neutral passthrough contract: ordered image, video, and audio reference URLs plus prompt and output settings.

How to get better Seedance 2.5 results

• Give every reference an explicit role in the prompt with @name, and describe what each one controls - subject, scene, motion, style, voice, or rhythm.
• For long clips, write beats in chronological order with timestamps so the model can map story, camera, and audio to the right seconds.
• Start with fewer references and add assets only when each one adds information the prompt cannot express.
• Repeat stable visual anchors (clothing, silhouette, lighting, material) when several references must read as one character or product.
• For editing or extension, set Aspect ratio to Adaptive. Editing also requires Duration to Match source; fixed settings can be rejected after Seedance classifies the prompt.
• Review motion physics, multi-subject interaction, and audio sync before using a result in a longer sequence.

Limits and availability

• MachGen admits only the modes and settings listed above for this catalog model.
• Source assets must finish uploading before a request can be submitted.
• Generated results can vary between requests, including when the same prompt and settings are reused.
• Seedance 2.5 is pass-through. MachGen sends the request through its upstream provider integration and returns the generated asset.

Learn more

For model background, technical details, and architecture, see the official Seedance 2.5 page.

Try this model in the MachGen Playground.

Official documentation may describe capabilities or parameters that are not currently exposed by MachGen. Use the Inputs and output settings section above as the source of truth for this page.

Seedance 2.5 pricing

See Pricing for currently published MachGen rates. When available, the Playground estimate reflects the selected task and output settings.

Seedance 2.5 on MachGen

Deployment: Pass-through. Seedance 2.5 is pass-through. MachGen sends the request through its upstream provider integration and returns the generated asset.

The Playground and API sections on this page list the inputs and output settings exposed for this model.