- How LTX API pricing works: usage-based pricing determined by endpoint, model, resolution, and duration.
- How to estimate spend with a simple formula and worked example, plus Fast vs Pro cost trade-offs.
- How automatic duration affects prepaid billing, and how to control API spend.
LTX API pricing is usage-based, with rates determined by the endpoint, model, resolution, and duration being billed. Text-to-video and image-to-video are billed by generated video duration, while audio-to-video is billed by input audio duration and editing endpoints have their own billing rules. This guide explains how LTX API pricing works and how to turn current rates into a reliable estimate of your API spend.
If you are integrating an AI video generation API into a product, the pricing model matters as much as output quality because it determines your unit economics at scale. LTX publishes per-second rates by endpoint, model, and resolution, making it possible to estimate generation costs directly from the workload you plan to run.
How does LTX API pricing work?
LTX API pricing is generally billed per second, with the billing basis depending on the endpoint. Text-to-video and image-to-video are billed by generated video duration, while audio-to-video is billed by input audio duration. Editing endpoints have their own billing rules.
The per-second rate depends on the endpoint, model, and output resolution. Higher resolutions generally cost more per second, and Pro costs more than Fast at the same supported resolution.
What is a pricing SKU?
A SKU is the specific combination of endpoint, model, resolution, and output type that a request resolves to. Each SKU has its own per-second rate. For example, a SKU is written in the form image-to-video-ltx-2-5-pro-1920X1080-audiovisual, which encodes the image-to-video endpoint, the ltx-2-5-pro model, a 1920x1080 resolution, and audiovisual output. When you change any one of those, you move to a different SKU with its own price.
Reading the /pricing endpoint
The /pricing endpoint returns the full pricing schedule for every SKU, and it does not require authentication. You can call it directly:
curl --location 'https://api.ltx.video/pricing'
The response maps each SKU to an array of { price_cents, effective_date } entries. A static-price SKU has a single entry with effective_date: null. A SKU with a scheduled change has a base entry plus dated entries for upcoming prices. Because a SKU can contain dated pricing entries, rates can change on a schedule. Use the /pricing endpoint as the source of truth for the current rate rather than hardcoding a price into your application.
Fast vs Pro: cost per model
Model is one of the main levers on your per-second rate. For LTX-2.5, the API offers ltx-2-5-fast for speed and lower cost and ltx-2-5-pro for higher fidelity. Pro costs more per second than Fast at the same supported resolution.
For workloads where speed and cost dominate, such as previews, drafts, or high-volume generation, the Fast tier keeps the per-second rate down. For final renders where fidelity is the priority, the Pro tier is the higher-cost option. You can use both within a single application: iterate on Fast, then use Pro for final output, with each generation billed at the rate for the model and resolution you select. If you are unsure which tier fits a given job, the model selection guide walks through the decision.
How does resolution affect cost?
Resolution is another lever on your per-second rate. Higher resolutions cost more per second than lower ones on the same model and endpoint, because resolution is part of the SKU. A request at 1920x1080 maps to a different, higher-priced SKU than the same request at a lower resolution. For example, LTX-2.5 Fast text-to-video currently costs $0.09 per second at 720p, $0.13 at 1080p, $0.19 at 1440p, and $0.30 at 4K.
You set resolution with the resolution field, for example "1920x1080". The practical takeaway for cost control is to render at the resolution you actually ship at, and to use lower resolutions for iteration passes where final fidelity does not matter yet. Since duration multiplies the per-second rate, resolution and duration together are the two numbers that most directly shape a clip's cost.
How to estimate your spend
Estimating spend on the LTX API comes down to three inputs per workload: the per-second rate, the applicable duration for the endpoint, and how many generations you make. For text-to-video and image-to-video, that means generated clip duration; for audio-to-video, it means input audio duration.
The cost formula
The cost formula is: cost per clip in cents equals the SKU per-second rate multiplied by the clip duration in seconds. Monthly spend equals cost per clip multiplied by the number of clips per month. Written out:
clip_cost_cents = price_cents_per_second × duration_seconds monthly_cost_cents = clip_cost_cents × clips_per_month
A worked example
A worked example makes the formula concrete. Take LTX-2.5 Fast text-to-video at 1920x1080, currently priced at $0.13 per second. A 6-second clip costs $0.13 × 6 = $0.78. If your application generates 1,000 such clips a month, the estimated generation cost is $780. Switch those same clips to a different model or resolution and recompute using the current rate for that configuration.
Prepaid vs postpaid billing
Prepaid and postpaid are the two billing types on the LTX API, stored per organization. Prepaid accounts, created automatically on developer console signup, pay before use by topping up a credit balance with a card. Postpaid accounts, set up manually for enterprise customers with contracts, are invoiced after use with no balance to maintain.
Prepaid credits are tracked as one-cent tokens, so a 10-dollar top-up gives you 1,000 tokens. On a prepaid call, the API uses a reserve-then-charge pattern: it reserves the estimated cost before generation, generates the video, then finalizes the charge on success or releases the reservation on failure. You are only charged for successful generations, an unfinalized reservation expires after 15 minutes, and if your balance is too low the request returns 402 Payment Required. For teams running video generation at scale, postpaid invoicing removes the need to keep a balance topped up.
LTX-2.5 automatic duration adds one consideration for prepaid accounts. Because the final clip length is not known before generation, credits are temporarily held against the longest duration allowed by the selected resolution and frame rate. Depending on the configuration, that can mean covering up to 20 seconds for ltx-2-5-fast or 10 seconds for ltx-2-5-pro. Once generation finishes, the account is charged for the length actually generated and the unused remainder is released. Postpaid accounts do not place this hold and are invoiced for the generated length.
How can you control API spend?
Controlling API spend starts with choosing the right model, resolution, and duration for the workload. Choose the Fast tier wherever its fidelity is sufficient, and reserve the Pro tier for final renders, since Pro costs more per second. Right-size resolution and duration to what you actually ship, because both multiply into the final charge.
Beyond the per-request choices, read the /pricing endpoint at integration time and cache it briefly rather than hardcoding rates, so scheduled price changes flow through automatically. Track the sku and total_billed_cents fields that every successful call logs, and you have a per-SKU breakdown of exactly where your spend goes. For a broader view of what AI video generation costs across surfaces, see the cost of AI video generation.
Summary
LTX API pricing is generally billed per second, with the billing basis depending on the endpoint and rates varying by model and resolution. Text-to-video and image-to-video are billed by generated video duration, while audio-to-video is billed by input audio duration. Fast provides the lower-cost generation path, while Pro costs more per second for higher-fidelity output. Editing endpoints have their own billing rules and model availability; current Retake, Extend, HDR Upscale, and Video Reframe pricing is listed under LTX-2.3 Pro. To estimate spend, use the current rate for your endpoint, the applicable duration, and your expected generation volume.
Ready to build? Get an LTX API key and check the live rates on the LTX API pricing page.
