🚀 One API for every frontier model on FastInfra

Browse models

Transparent per-token pricing → compare providers on FastInfra

View pricing

OpenAI-compatible chat completions → start in minutes

Read the docs

⚡ Server Auction → Enterprise bare metal from $78.15/mo (173 in stock)

Browse deals

Video API · LTX-2.5 · Self-hosted on FastInfra H100s

Text-to-video API on FastInfra GPUs

The FastInfra text-to-video API renders LTX-2.5 clips with synced audio on hardware we run ourselves and bills $0.10 per second of finished video, with nothing charged for a failed render. Submit a prompt and a duration between 2 and 20 seconds, poll the job, and download an MP4, all with the same API key, credits and invoice you use for chat and speech. Live counts of rendered videos come from the billing roll-up and the renderer's status is published on the status page.

Text-to-video with synced audio through the same OpenAI-style key you use for chat. Submit a prompt, poll the job, download the MP4. $0.10 per second of output, billed only for completed clips.

  • Runs on our own GPUs
  • No training on your prompts
  • Async jobs, pay on completion
Create free account Video API docs Video renderer: operational

100.0% uptime over 299 probes in 24 h · median probe 419 ms

Measured, not estimated

99,282 Videos rendered completed jobs since 26 Aug 2026
853,328 Seconds of video billed output, lifetime
33,645 Last 30 days completed · 97.54% of jobs finished without failing

Counted from the billing roll-up, not from the pod's health counter (which resets on restart). Refreshed every ten minutes.

Reliability: One H100 renders every job today; the status page shows its 24-hour uptime and queue depth. A second renderer is planned before we quote an uptime number here.

One key, one bill

Video jobs share the API key, credits and invoice with chat, embeddings and speech. No separate account, no per-seat plan.

Priced per second, not per credit

$0.10 per second of finished video at 1280x704. A failed render is not billed. The prompt costs $0.10 per million tokens, which rounds to nothing.

Our hardware, our queue

LTX-2.5 distilled runs on a FastInfra H100, not a marketplace. Queue depth and readiness are published live on the status page.

How video is priced

What Price Token equivalent Notes
Output video $0.10 per second 1,000 output tokens per second at $100.00 / 1M Default 1280x704; 1536×1024 accepted via size
Prompt $0.10 per 1M tokens ≈ 1 token per 4 characters Effectively free

Worked example

  • 5-second clip (default)$0.50
  • 10-second product shot$1.00
  • 20-second clip (maximum)$2.00
  • 1,000 five-second clips$500.00

Prices are USD, prepaid credits, no platform fee. The same numbers appear on the pricing page under Video.

Code

# 1. submit the job (HTTP 202)
curl https://api.fastinfra.ai/v1/videos \
  -H "Authorization: Bearer $FASTINFRA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "lightricks/ltx-2.5",
    "prompt": "A slow dolly shot over a misty pine forest at sunrise, birds taking off",
    "seconds": 8,
    "size": "1280x704"
  }'

# 2. poll until status is "completed"; the MP4 is in data[0].b64_json
curl https://api.fastinfra.ai/v1/videos/$JOB_ID -H "Authorization: Bearer $FASTINFRA_API_KEY" \
  | jq -r '.data[0].b64_json' | base64 -d > clip.mp4
import base64, time, requests

H = {"Authorization": f"Bearer {API_KEY}"}
job = requests.post("https://api.fastinfra.ai/v1/videos", headers=H, json={
    "model": "lightricks/ltx-2.5",
    "prompt": "A slow dolly shot over a misty pine forest at sunrise",
    "seconds": 8,
}).json()

while job["status"] not in ("completed", "failed"):
    time.sleep(5)
    job = requests.get(f"https://api.fastinfra.ai/v1/videos/{job['id']}", headers=H).json()

if job["status"] == "completed":
    open("clip.mp4", "wb").write(base64.b64decode(job["data"][0]["b64_json"]))
import fs from "node:fs";
const H = { Authorization: `Bearer ${process.env.FASTINFRA_API_KEY}`, "Content-Type": "application/json" };
let job = await (await fetch("https://api.fastinfra.ai/v1/videos", {
  method: "POST", headers: H,
  body: JSON.stringify({ model: "lightricks/ltx-2.5", prompt: "A slow dolly shot over a misty pine forest", seconds: 8 })
})).json();

while (!["completed", "failed"].includes(job.status)) {
  await new Promise(r => setTimeout(r, 5000));
  job = await (await fetch(`https://api.fastinfra.ai/v1/videos/${job.id}`, { headers: H })).json();
}
if (job.status === "completed") fs.writeFileSync("clip.mp4", Buffer.from(job.data[0].b64_json, "base64"));

Volume

Pay as you go

Any volume

$0.10 / s

Prepaid credits, no commitment, no minimum.

Committed volume

From $500 / month

Quoted

Tell us your monthly seconds and we quote a committed-use rate; reply comes from the support queue, not a sales funnel.

Dedicated renderer

Your own H100 pod

$3.90 / hr

Rent an H100 on GPU Cloud and run the same LTX image yourself: unlimited renders for the hourly rate; at ~60 s per 5 s clip that is roughly $0.013 per second if the pod stays busy.

Committed-volume requests go through the contact form and land in the same support queue as everything else. Contact form.

Compared with

Provider Published price Notes
fal.ai LTX-2.5 fast $0.09 / s at 720p · $0.13 / s at 1080p Per-second list price; 1280×704 output sits between the two
Replicate Per-second GPU time You pay for the whole render including retries; FastInfra bills only completed output

Competitor prices are their published list prices at the time of writing and may change; check their pages before deciding on price alone.

Limits and formats

  • Duration 2–20 seconds per clip (default 5).
  • Default size 1280x704; width and height must be divisible by 64. 1536×1024 renders about 1.8× slower.
  • Output is MP4 with synced audio, returned base64-encoded in data[0].b64_json; the job keeps the prompt and result in your job history until it expires.
  • Asynchronous only: submit returns 202 and a job id; poll or use the console to download.

Where it runs and what we keep: see regions and data handling.

Frequently asked questions

How much does the text-to-video API cost?

$0.10 per second of finished video at 1280×704, so a five-second clip is $0.50 and a twenty-second clip is $2.00. The prompt is billed at $0.10 per million tokens, which rounds to nothing, and a render that fails is not billed at all.

Which model does the video API use?

LTX-2.5 distilled by Lightricks, served from a FastInfra H100 rather than a third-party marketplace. Request it as lightricks/ltx-2.5 or the aliases ltx-2.5, ltx-2.5-distilled and ltx-2-5-fast; the output includes synced audio.

Is video generation synchronous?

No. POST /v1/videos returns 202 with a job id; poll GET /v1/videos/{id} until the status is completed and read the MP4 from data[0].b64_json. A five-second clip typically renders in under a minute; 1536×1024 takes about 1.8 times longer.

What are the video limits?

Clips run from 2 to 20 seconds, the default is 5. Width and height must be divisible by 64; 1280×704 is the default and 1536×1024 is accepted. Jobs and their results stay in your history until they expire, and prompts are never used for training.

Can I run the video model on my own GPU instead?

Yes. Rent an H100 pod on GPU Cloud and run the same LTX image; you pay the hourly GPU rate with unlimited renders, which beats per-second pricing once the pod is busy most of the hour. The volume section on the page shows the live hourly rate.

Reviewed by Ashish Kumar, Head of Product · Updated November 17, 2026. Prices and counts on this page come from the live catalog.

Render your first clip in a minute

Create a free account, add credits and post a prompt. A 5-second clip costs $0.50; a failed render costs nothing.