Video generation
26 video models, one API key, real prices
Google Veo 3.1, Kling 3 Pro with motion control, Seedance 2.0, Hailuo-02, Grok Imagine Video and more — behind the same ak- key you already use for chat, with the exact USD cost shown before you generate.
Text-to-video, image-to-video, more
26 video models today, from Google, Kuaishou, ByteDance, MiniMax, xAI and more. Full list at /models, live from GET /v1/models.
First frame, last frame, references
Guide a shot with a first and/or last frame, or give Seedance 2.0 up to 9 reference images and address them right in the prompt as @Image1.
Motion control
Kling 3 Pro can transfer the motion of a reference clip onto your character — you keep the character, it keeps the movement.
Duration, aspect, resolution, audio
Controls per model — the UI only offers what the selected model actually accepts, so you don't discover a limit from a 400 error.
Async jobs, real prices upfront
Video runs as a background job: POST returns 202, close the tab, the result lands in your gallery. The composer shows the exact USD price first — a 4s 720p Veo 3.1 clip is about $1.44 — and failed jobs are never charged.
Dashboard, API, or MCP
Use the visual studio at /videos, call the async API directly, or reach it from the apexapi-mcp server (generate_video tool) and the n8n community node. Guardrails and rate/concurrency limits apply everywhere.
The flagships
26 video models total. These are the ones people ask for by name.
Call it directly
Async, OpenAI-shaped: start the job with POST /v1/videos/generations, then poll it.
# Start a video job (returns 202 + job id)
curl https://api.apexapi.dev/v1/videos/generations \
-H "Authorization: Bearer ak-your-key" \
-H "Content-Type: application/json" \
-d '{
"model": "google/veo3.1",
"prompt": "a donkey slowly turning its head toward the camera, golden hour",
"duration": 4
}'
# Poll it
curl https://api.apexapi.dev/v1/videos/generations/JOB_ID \
-H "Authorization: Bearer ak-your-key"Video generation FAQ
How many video models are available?
26 video models in the live catalog today, out of 139 models total across the API. The full list with pricing is at apexapi.dev/models, always current from the same data behind GET /v1/models.
How much does a video cost?
It depends on the model, duration, and resolution. As an example, a 4-second 720p clip on Google Veo 3.1 is about $1.44. The composer at /videos shows the exact USD price for your current settings before you submit the job, and failed jobs are never charged.
Why does video generation return a job ID instead of the result?
Video takes longer than a request-response cycle can comfortably hold open, so POST /v1/videos/generations returns 202 with a job ID and you poll GET /v1/videos/generations/{id}. In the dashboard you can just close the tab — the finished clip lands in your gallery.
Can I control how the video starts and ends?
Yes. Most models accept a first frame and/or a last frame to anchor the shot. Seedance 2.0's reference-to-video mode goes further: up to 9 reference images you can address directly in the prompt as @Image1, @Image2, and so on.
What is motion control?
On Kling 3 Pro, motion control transfers the movement from a reference clip onto your own character or subject — you supply the character, the reference clip supplies the motion.
Is there an API, not just the dashboard?
Yes — the same ak- key you use for chat and images. POST /v1/videos/generations starts the job and GET /v1/videos/generations/{id} polls it. It's also the generate_video tool on the apexapi-mcp MCP server and available through the n8n community node.
Generate your first clip
Open the studio, pick a model, and see the price before you spend it — on ApexApi.