AI video generation API: settings, cost and wait times

The Video Studio in your apillms console turns a text prompt — or a start image — into a video clip. This post walks through every setting, shows what each model costs per clip, covers common issues, and explains what happens if a job fails.

Step-by-step: generating your first clip

1. Log in at /login and open Video Studio from the console menu.
2. Pick a model from the dropdown. Each model exposes its own resolution, duration and aspect ratio options.
3. Write a prompt describing the shot: subject, action, camera movement, lighting. The more specific, the more predictable the result.
4. Choose resolution, duration and aspect ratio. Watch the estimated cost update as you change these.
5. If the model supports it, toggle generate audio on if you want a soundtrack or sound effects.
6. Optionally paste a start image URL to anchor the first frame to an existing image instead of generating purely from text.
7. Review the cost estimate one more time, then click generate.
8. Wait — video jobs can take several minutes. You can close the tab and check the Media library later.

Choosing a model

Video models are billed in credits, where 1 credit = $0.005. Here's the cheapest-tier cost per clip for each model currently on apillms:

These are cheapest-tier figures; the actual cost shown in the Studio depends on the resolution, duration and aspect ratio you pick, since higher settings use more credits. Always check the estimated cost on screen before generating — it reflects the exact combination of options you've chosen, not the base figure in the table.

If you're unsure which model fits a given shot, a practical approach is to test the same prompt on two cheap models first (say veo3.1-lite and kling-2.5-turbo) before committing credits to a more expensive one like kling-o1-video. Check /models for the full current list with capability filters, and /compare if you also want to see how apillms model pricing compares across chat/image/video categories.

Resolution, duration and aspect ratio

Each video model exposes its own set of resolution and duration options in the Studio — pick what the clip actually needs rather than defaulting to the highest tier. Higher resolution and longer duration increase the credit cost, and the Studio recalculates the estimate live as you adjust them. Aspect ratio (landscape, portrait, square, depending on the model) is a separate setting and also affects which resolution/duration combinations are available.

A practical workflow for controlling cost:

• Start with the shortest duration and lowest resolution to validate the prompt and framing.
• Once the motion and composition look right, re-run at the resolution and duration you actually need for the final clip.
• If the model offers multiple aspect ratios, decide on the target platform (landscape for most uses, portrait for short-form vertical video) before generating, since switching aspect ratio later usually means regenerating the whole clip.

Generating audio and using a start image

Some video models support a generate audio toggle, adding a soundtrack or sound effects to the clip. Turn it on in the Studio if the model supports it; the cost estimate updates accordingly once enabled, so compare the with-audio and without-audio estimates before deciding.

You can also provide a start image URL instead of generating purely from text. This anchors the first frame of the video to that image, which is useful when you want to:

• Animate a still image you already made — for example in the Image Studio.
• Keep a consistent character or scene across several clips by reusing the same still as the starting frame.
• Turn a product photo or logo into a short motion clip without redrawing the subject from scratch.

The image needs to be reachable at a public URL — if you generated it in Image Studio, you can grab its URL from the Media library or directly after generation, then paste it into the Video Studio's start image field.

Cost shown before you generate

The Video Studio always shows the estimated cost for your current settings before you submit the job. Review it, especially after changing resolution or duration, since those are the variables that move the price the most. Your credits come from whichever pack you've purchased — Starter ($5 / 1000 credits), Builder ($20 / 4200 credits), Pro ($50 / 11000 credits) or Scale ($100 / 23000 credits) — and credits never expire, so you can top up once and use them across image, video and voice generation at your own pace. See exact token and credit breakdowns on /pricing, and payment history on the Billing page.

Wait times and failed jobs

Video generation is not instant — jobs can take several minutes to complete, depending on the model and settings. You don't need to keep the Studio tab open and waiting; check back once the job finishes.

If a job fails, it's refunded automatically — you are not charged for failed video generations, the same policy that applies to failed API requests across apillms. Completed videos land in your Media library alongside generated images and audio, but they're kept only for a limited time, so download anything you want to keep rather than relying on the library as long-term storage.

If a clip doesn't match your prompt well, generating failed isn't usually the issue — a completed but unsatisfying clip is still billed, since it succeeded as a job. In that case, adjust the prompt (be more specific about camera angle, motion, and subject) or try a different model rather than regenerating the same settings repeatedly.

Triggering video generation from an agent or script

If you'd rather not use the console UI every time, the apillms MCP server exposes the same video functionality as tools: generate_video, list_video_models and job_status, alongside tools for chat, images and voice. Connect it to Claude Code with:

claude mcp add --transport http apillms https://apillms.com/mcp --header "Authorization: Bearer YOUR_KEY"

Once connected, you can ask your agent to list available video models, kick off a generation with a prompt and settings, and poll job_status until the clip is ready — useful if you want video generation as a step in a larger coding or content workflow rather than a one-off console task. See the full walkthrough in the MCP server guide, which also covers the chat and image tools available over the same connection.

Takeaway

Pick a model, set resolution/duration/aspect ratio, optionally add audio or a start image, and check the live cost estimate before generating. Test cheap and short first, scale up once the prompt is dialed in, expect a wait of several minutes per clip, and know that failed jobs don't cost you anything — only completed-but-unsatisfying ones do, so iterate on the prompt before switching models.

FAQ

How much does an AI video clip cost on apillms?

It depends on the model and settings, but cheapest-tier clips range from $0.16 (veo3.1-lite) to $1.38 (kling-o1-video) per clip, based on credits at $0.005 each. The exact cost for your chosen resolution, duration and aspect ratio is always shown before you generate.

How long does video generation take?

Video jobs can take several minutes to complete. You don't need to keep the page open — check the Media library once it's done.

What happens if my video generation job fails?

Failed jobs are refunded automatically, so you're not charged for them. A completed clip that just doesn't match your prompt well, however, is still billed since the job succeeded.

Can I animate an existing image instead of starting from text?

Yes — paste a start image URL in the Video Studio to anchor the first frame of the clip to that image, including images you generated in the Image Studio.