Seedance 2.5 — Text, Photo, Video & Audio to Video
Seedance 2.5 is ByteDance's video generation model, available through reAPI's unified endpoint and designed to produce 4–30 second video clips in a single pass from text, images, video, and audio references. It can render in 480p, 720p, or 1080p and includes audio output synchronized to the visuals, with support for speech, sound effects, and music. Requests are sent to an asynchronous API, and the generation mode is selected implicitly by the media fields included in the request. The model exposes five generation modes: text-to-video, image-to-video, first-frame/last-frame interpolation, multimodal reference generation (up to 30 images, 10 video clips, and 10 audio tracks), and prompt-driven video editing. In addition to standard between-frames and image-reference use, it accepts audio-only input, and editing durations are controlled with the duration parameter set to -1 to preserve the source clip length. Users can select aspect ratios such as 9:16, 1:1, and 16:9, choose MP4 or MOV output, optionally disable watermark, and request the last frame URL for chaining clips. Seedance 2.5 is particularly suited to paid workflows at scale because pricing is per second of actual generated output. On reAPI the model is promoted with a rate starting at $0.071 per second and supports 4–30 second clips, meaning a single request can cost roughly $0.28–$2.13 depending on duration and resolution. This makes abandoned or low-quality takes relatively inexpensive to iterate on.
- Category
- AI Video Generator
- Pricing
- Unknown
Problem solved
Seedance 2.5 solves the difficulty of producing short-form video with synchronized audio without a full production pipeline. It replaces multiple specialized tools by letting users pass text prompts, still images, video clips, reference footage, and audio tracks to a single asynchronous endpoint. Per-second billing and automatic duration selection reduce wasted spend, while real-person references and multimodal inputs support consistent characters and context-aware generation for commercial video.
Use cases
- Create product demo clips from a single product photo
- Turn still images into social videos for Reels, Shorts, and TikTok
- Generate talking-head clips with synced speech
- Interpolate between first and last frames for seamless motion
- Combine up to 30 reference images, 10 video clips, and 10 audio tracks for guided generation
- Perform prompt-driven video editing while preserving source clip length
Audience
- Developers
- Content Creators
- Product Managers
- Startups
- Enterprise
Pros
- Creates 4–30 second videos with audio in one pass
- Supports text, image, video, and audio inputs on a single endpoint
- Multiple generation modes, including first/last frame and multimodal references
- Per-second pricing and auto-duration control reduce wasted spend
- Output options include 480p/720p/1080p, MP4/MOV, watermark toggle, and last-frame chaining
Cons
- No self-hosting; requires an API subscription through a provider like reAPI
- Pricing is per second and not published as a fixed flat rate across providers
- 30-second 1080p generations can become expensive at scale
- Model architecture and roadmap are not open
- Async processing requires polling or webhooks rather than synchronous response
Official website: https://reapi.ai/models/seedance-2-5
Canonical LambdaBase page: https://www.lambdabase.com/tools/seedance-2-5-text-photo-video-audio-to-video