Video Generation API
Generate short-form video from text or image inputs through one production-ready API. Access Seedance 2.0, Wan 2.7, Kling, Hailuo, and more — one key, async job handling, unified pricing.
Production video routing for text-to-video and image-to-video.
Keep one async workflow while you choose between cinematic quality, preview speed, real-person review, and cost tier.
Typical completion
Async job orchestration built in for heavyweight generations.
Delivery tiers
Choose preview speed or commercial output without changing vendors.
Generation modes
Text-to-video and image-to-video share one contract.
Overview
Why video generation needs a routing layer
Video generation models are harder to operate than image or chat models. Generation takes 15–120 seconds. Providers have inconsistent rate limits, output formats, and pricing units. A single provider may excel at cinematic quality but fail on fast-preview turnaround.
Why video generation needs a routing layer
Video generation models are harder to operate than image or chat models. Generation takes 15–120 seconds. Providers have inconsistent rate limits, output formats, and pricing units. A single provider may excel at cinematic quality but fail on fast-preview turnaround.
Every major video lab ships updates on its own cycle. A model that was average last month may become best-in-class today — and one that led the benchmark may degrade after a silent update. With a single provider integration, your application has no way to respond. With ImaRouter, swapping the underlying model is one parameter change, not a re-integration.
ImaRouter standardizes the integration layer so your application sends one request with your prompt, input image (if needed), duration, and quality preference — and the API handles provider selection, async job management, and output delivery. You focus on the user experience; we handle the infrastructure.
- One API for text-to-video and image-to-video across all top providers
- Built-in async job handling — no need to build your own polling infrastructure
- Provider routing based on speed tier or quality tier
- Switch models instantly when a provider releases a better update — no re-integration needed
- Consistent output format: MP4 via CDN URL across all models
- Real-person video generation with built-in compliance review layer
Supported models and generation modes
ImaRouter integrates the leading video generation models across two generation modes: text-to-video and image-to-video. Each model has a distinct profile for quality, speed, style, and supported duration.
- Seedance 2.0 — Cinematic quality, 720p/1080p, max 15s, fast and standard tiers, real-person support
- Wan 2.7 — High motion coherence, strong on narrative scenes, multilingual prompt support
- Kling AI — Photorealistic output, strong camera control, suitable for product and lifestyle
- Hailuo — Fast preview tier, good for storyboard iteration and UGC content
- Vidu — Avatar-style and portrait video, good for talking-head generation
- Pixverse — Creative and stylized output, ideal for social media short-form
Capabilities
How async video generation works
Video generation is always asynchronous. A single generation request kicks off a job that typically completes in 15–90 seconds depending on model, resolution, and duration. ImaRouter abstracts the full async lifecycle.
How async video generation works
Video generation is always asynchronous. A single generation request kicks off a job that typically completes in 15–90 seconds depending on model, resolution, and duration. ImaRouter abstracts the full async lifecycle.
Submit a generation request and receive a job ID immediately. Poll the job status endpoint or configure a webhook URL to receive the completed video. The response includes the output video URL, model used, generation time, and cost in USD.
- POST /v1/video/generate — submit job, returns {jobId, status: 'queued'}
- GET /v1/jobs/{jobId} — poll for status: queued → processing → completed
- Webhook: set a callback URL and receive the completed payload with no polling
- Failed jobs return structured error with reason code and retry guidance
What you can build
Video generation at API scale unlocks product patterns that weren't economically or technically feasible before. These are the integration patterns teams are using in production today.
- Ad creative generation: produce dozens of video variants from a single product image for A/B testing
- UGC tools: let users generate short personalized clips from prompts inside your product
- Storyboard automation: convert script lines to rough video clips for pre-production review
- E-commerce product video: animate product stills into lifestyle video for listing pages
- Social content scheduling tools: generate and schedule video content in one automated pipeline
- Real-person video: produce spokesperson clips backed by ImaRouter's human review compliance layer
Quality tiers and pricing
Video generation cost varies by model, resolution, and duration. ImaRouter surfaces two quality tiers across all models to help you manage cost at scale.
The fast tier prioritizes turnaround time and is suited for previews, iteration, and UGC. The standard tier prioritizes output quality and is suited for commercial delivery. You can switch tiers with a single parameter change — no separate API keys or accounts needed.
- Fast tier: optimized for speed, lower cost, ideal for preview loops and UGC
- Standard tier: optimized for quality, commercial-grade output
- Pricing from $0.18 per generated second (Seedance 2.0 fast tier)
- No platform fee or minimum commitment — pay per generated second
- Cost visible per job in response body and in dashboard usage logs
FAQ
FAQ
What is the maximum video length I can generate?
Maximum duration depends on the model. Most models support up to 10–15 seconds per generation request. Wan 2.7 supports up to 20 seconds. For longer videos, generate multiple clips and stitch them in post — the ImaRouter response includes the CDN URL for each clip.
Can I generate video from an image (image-to-video)?
Yes. All major models support image-to-video generation. Pass the input image URL in the source_image field of the request along with your motion prompt. The model will animate the image according to the prompt, preserving the composition and subject.
How do I handle real-person video generation?
ImaRouter routes real-person video requests through a human review layer before delivery. This adds 1–4 hours to delivery time but ensures output meets commercial compliance standards. Pass the flag review_required: true in your request to enable this flow.
What resolution and format are videos delivered in?
All videos are delivered as MP4 files via CDN URL in the job completion response. Supported resolutions are 720p and 1080p depending on model and tier selected. 4K output is available on select models — see the pricing page for model-level specs.
How long do generated videos stay available on the CDN?
Generated video files are available at the CDN URL for 72 hours after generation. Download and store files in your own storage before this window closes. The dashboard shows download links for all recent jobs.
Launch paths
Related links and launch paths
More modalities
Explore more routes from the same stack
Text-to-Image API
Generate stunning visuals through a unified API layer. Top models: GPT Image 2, Midjourney, Nano banana Pro — all via one clean interface.
Learn more
LLM Routing API
One key for every frontier and open-weight LLM. Route Claude, GPT, Deepseek, and Qwen with automatic fallback, cost caps, and unified billing — zero per-provider wiring.
Learn more
Speech & Audio API
Transcribe with Whisper, synthesize voice with ElevenLabs, generate music with Suno — all through one endpoint. Same key, same billing as your video and image workloads.
Learn more