SEEDANCE 2.5 · LIVE ON ANYCAP
AnyCap schema checked September 9, 2026
Seedance 2.5
API for AI agents
Use seedance-2.5 through AnyCap to generate video from text, images, first and last frames, or multimodal references. Choose a supported mode, inspect its schema, and save the clip into your agent workflow. The AnyCap request limits and examples come first below; Dreamina product features are documented separately so you can tell which controls apply to your request.
AnyCap availability and verified modes
AnyCap availability
Live
Schema checked
September 9, 2026 · CLI 0.6.1
AnyCap CLI model ID
seedance-2.5
Verified modes
text-to-video, image-to-video, first-last-frame-to-video, multi-modal-reference
Pricing
Check the live AnyCap catalog
How to use Seedance 2.5 on AnyCap
Start with discovery instead of copying provider-specific request fields. AnyCap keeps authentication, model discovery, execution, and output delivery in one agent-facing workflow while Seedance 2.5 performs the video generation.
Discover the model
anycap video modelsInspect the mode you will use
anycap video models seedance-2.5 schema --operation generate --mode image-to-videoGenerate from text
anycap video generate --model seedance-2.5 --mode text-to-video --prompt "a cinematic product launch shot with controlled camera motion" --param duration=5 --param resolution=720p -o seedance-2-5.mp4Animate your own product image
anycap video generate --model seedance-2.5 --mode image-to-video --prompt "a slow camera move around the product" --param images='["./product.png"]' --param duration=5 --param resolution=720p -o product-video.mp4AnyCap input fields, limits, and output
These values come from a read-only schema check on September 9, 2026. The generation commands above use verified field names; replace product.png with your own image. No generated-video quality claim is implied by a catalog or schema check.
| Mode | Input fields | Use and limits |
|---|---|---|
| text-to-video | prompt | Generate from a text brief; no reference asset is needed. |
| image-to-video | images, videos | The schema allows up to 9 images and 3 videos. Use your own local files or HTTP(S) URLs. |
| first-last-frame-to-video | first_frame, last_frame | Provide the named start and end images for frame-to-frame generation. |
| multi-modal-reference | images, videos, audios, prompt | Up to 9 images, 3 videos, and 3 audio references. Explain what the model should inherit from each input. |
duration: 5–30 seconds. In multimodal mode with reference video, use-1to match the complete reference-video duration.resolution: 480p, 720p, or 1080p. The checked AnyCap schema does not offer 4K.aspect_ratio: choose an enumerated ratio. In multimodal mode with reference video, useadaptiveto follow that video.generate_audiocontrols generated audio; it is separate from audio reference inputs.
Continue with the returned video
The CLI normally waits for completion and downloads the video to the path selected with -o. Inspect the returned status and local file before referencing it in a page or sharing it. With --no-wait, continue tracking the returned task rather than assuming the output file already exists. See the CLI reference for task and output handling.
Before submitting an AnyCap generation
CHECK 1
Discovery exposes the active seedance-2.5 runtime identifier.
CHECK 2
The live catalog declares text, image, first/last-frame, and multimodal-reference modes.
CHECK 3
Agents inspect the current schema before sending local image, video, or audio references.
CHECK 4
AnyCap keeps authentication, execution, task handling, and output delivery in one workflow.
Dreamina product reference: Seedance 2.5 vs 2.0
The clearest Seedance 2.5 upgrade is a larger multimodal reference budget and longer generation window. The comparison below is distilled from the supplied official product documentation; it describes the product workflow, not an AnyCap API contract.
| Capability | Seedance 2.0 | Seedance 2.5 |
|---|---|---|
| Reference images per request | Up to 9 | Up to 30 |
| Reference videos | Up to 3; 15 seconds total | Up to 10; 30 seconds total |
| Reference audio clips | Up to 3; 15 seconds total | Up to 10; 30 seconds total |
| Audio-only reference-to-video | Not supported | Supported |
| Direct generation duration | 4–15 seconds | 4–30 seconds |
Dreamina reference-input limits
Seedance 2.5 can use image, video, and audio references in the same creative workflow. Dreamina's public page describes up to 50 combined multimodal inputs; the supplied operating documentation breaks that budget down by asset type. These limits are useful for planning source assets but are not an AnyCap API contract.
JPEG, PNG, WebP, BMP, GIF, HEIC, HEIF
Images
Up to 30 reference images
Each image can be 300–6,000 px with an aspect ratio from 0.4 to 2.5.
MP4, MOV
Video
Up to 10 clips and 30 seconds total
Reference video accepts 480p–4K input at 24–60 FPS; each clip can be 2–30 seconds.
WAV, MP3
Audio
Up to 10 clips and 30 seconds total
Each reference clip can be 2–30 seconds, including audio-only reference workflows.
Dreamina duration and resolution
STANDARD GENERATION
4–30 seconds
The documented direct-generation range is 4–30 seconds, up from 4–15 seconds in Seedance 2.0.
LONG-VIDEO BETA
30–180 seconds
Dreamina describes one-pass long-video generation with continuity across longer scenes.
OUTPUT SURFACE
Verify resolution
Dreamina markets clean 4K output, while the supplied parameter sheet lists 480p and 720p generation settings. Resolution depends on the product surface and should not be assumed for an API.
How to use Seedance 2.5 on Dreamina
The model is now available on both Dreamina and AnyCap. Product controls can change, but the source-backed workflow is consistent: choose a mode, provide the references and instruction, set the available output controls, then review and refine the generated video.
STEP 1
Choose Seedance 2.5
Open video generation in Dreamina, select Seedance 2.5, then choose the generation or editing workflow that matches the source material you have.
STEP 2
Add only the references the shot needs
Upload text, image, video, or audio references within the documented limits. For multi-subject work, start with a smaller reference set before using the full input budget.
STEP 3
Set the output and write the instruction
Choose the available aspect ratio, resolution, and duration, then describe the subject, motion, camera behavior, audio, or localized edit in the prompt.
STEP 4
Generate, review, and refine
Review continuity, reference fidelity, motion, and audio alignment. Use extension or smart editing when the product surface exposes those workflows instead of restarting an otherwise usable result.
Dreamina product workflows
Multimodal reference generation
Seedance 2.5 expands the reference budget across images, video, audio, and text. The supplied documentation also describes first-frame, last-frame, and multimodal-reference inputs.
Long-video generation
The product workflow describes one-pass 30–180 second generation without manually stitching multiple short segments. This is a product-surface feature, not a verified AnyCap API mode.
Video extension
A source video of up to 30 seconds can be extended by 4–30 seconds per operation. Chained extensions can produce a video up to 60 seconds from a 30-second source.
Smart video editing
The documented editing workflow supports natural-language changes, region selection with masks, boxes, or points, and timestamp-aware instructions for localized edits.
Continue researching Seedance 2.5
Use these guides for reference planning, product workflows, and model selection. For actual AnyCap requests, use the live model discovery and schema commands on this page rather than copying product-surface parameters.
WORKFLOW GUIDE
How to use Seedance 2.5 on the current product surface
A detailed reference-planning and production workflow that agents can connect to the live Seedance 2.5 model through AnyCap.
Read the guide →
MODEL COMPARISON
Seedance 2.5 vs Kling, Veo, and Runway
Use the comparison when the decision is model fit rather than AnyCap availability. Verify live provider and runtime facts before implementation.
Read the guide →
Seedance 2.5 questions
Is Seedance 2.5 available now?
Yes. Seedance 2.5 is live through AnyCap with the model ID seedance-2.5. Agents should discover the model and inspect its current schema before submitting references.
What changed from Seedance 2.0 to Seedance 2.5?
Dreamina documents a larger reference budget and longer video workflows than Seedance 2.0. Those product limits do not define the AnyCap request schema. For AnyCap, inspect the selected mode and use the verified input limits listed above.
What inputs does Seedance 2.5 support?
The AnyCap schema checked September 9, 2026 exposes text-to-video, image-to-video, first-last-frame-to-video, and multi-modal-reference. Multimodal mode accepts up to 9 images, 3 videos, and 3 audio references. First/last-frame mode uses named first_frame and last_frame inputs.
How do I use Seedance 2.5?
Use AnyCap: run anycap video models to confirm availability, inspect the seedance-2.5 schema for the intended mode, then call anycap video generate with the verified fields and route the returned video into the next agent step.
Can an AI agent call Seedance 2.5 through AnyCap?
Yes. AnyCap exposes Seedance 2.5 as seedance-2.5 with text-to-video, image-to-video, first-last-frame-to-video, and multi-modal-reference modes.
What can an AnyCap user run instead?
Seedance 2.0 is the current nearby alternative in the AnyCap catalog. Its model page contains the verified ID, supported modes, schema workflow, and CLI examples.
Does AnyCap expose every Dreamina feature?
No. The current AnyCap schema defines the supported modes, fields, and output controls. Dreamina long-video, extension, smart-editing, and 4K product features are separate; do not send their parameters unless the selected AnyCap schema exposes them.
Source and update policy
AnyCap availability and request fields were checked with CLI 0.6.1 on September 9, 2026. Dreamina comparisons summarize the official product page and operating documentation reviewed August 7, 2026. Product controls and AnyCap request fields can differ. Recheck the live AnyCap schema before production and use the linked Dreamina source for current product-surface details.