Price Comparison of 19 Image and Video Generation APIs: As Low as $0.003 per Image

As AI-generated content moves from demo to production, API call costs directly determine product gross margin structures. A recent price survey covering 19 providers shows that for a single image labeled as 1K, API quotes range from $0.003 to $0.22—a gap of over 70x; for 720p video, the per-second price ranges from $0.025 to $1.50, a span of 60x. This is not simply a question of "who is cheaper," but a chaotic race involving billing units, parameter definitions, and documentation conventions.

Same 1K Label, Why Can Prices Differ by 70x

The huge price range stems first from the ambiguity of the "1K" label itself. Different vendors define 1K differently: some mean 1024×1024, some mean a long edge of 1024 pixels with a shorter other edge, and others tie 1K to specific aspect ratios. Different total pixel counts naturally make per-image costs incomparable.

Second, there is a fundamental difference in billing granularity. Some providers bill per "successful generation," with failed requests not charged; others bill per "API call," regardless of whether the output is usable. For scenarios requiring multiple retries to get a satisfactory result, the latter's actual cost can be far higher than the list price.

Third, low-price tiers often come with conditions. Quotes around $0.003 usually correspond to volume discounts, prepaid credits, or specific model versions, while standard on-demand pricing is noticeably higher. Therefore, when comparing across providers, it is necessary to distinguish between "entry price" and "regular price."

Billing per successful generation vs billing per API call: The former charges only when the model returns a usable result, with failed or timed-out requests not appearing on the bill; the latter charges as soon as each HTTP request reaches the server, regardless of whether the output is adopted. For generation tasks requiring multiple retries to obtain a satisfactory result, the actual cost of per-call billing can be several times the list price. When comparing prices, developers should request the provider's billing rules documentation, clarify whether failures, timeouts, content-filter blocks, and similar scenarios are charged, and incorporate the retry rate into the cost model.

The Unit Trap in Video Pricing: Per Second Is Not Per Second

The $0.025 to $1.50 per-second price band for 720p video looks clear but hides multiple variables that cannot be directly converted.

The core difference lies in the definition of "second." Some vendors bill by the duration of the generated video, while others bill by processing time or GPU occupancy time. The former is result-oriented, the latter resource-oriented, and with different model efficiencies the two can differ by several times.

The way resolution and frame rate are bundled also varies. Some 720p quotes default to 24fps, others to 30fps; the higher the frame rate, the greater the computation per unit of duration. In addition, whether an audio track is included, whether first/last frame control is supported, and whether tiered pricing by video length applies all change the final bill.

This means that multiplying Vendor A's "per-second price" by Vendor B's video duration to make a budget will likely lead to wrong conclusions. The truly comparable approach is to measure the total cost of a single video under the same resolution, frame rate, duration, and feature set.

Result-duration billing vs resource-duration billing: Result duration refers to the playback seconds of the final output video and is unrelated to model efficiency; resource duration is billed by GPU occupancy time or processing time, so the slower the model and the longer the queue, the higher the bill. The two may be close for efficient models, but can differ by several times in inefficient or high-load scenarios. In addition, some providers set tiered unit prices by total video length, with short clips costing more per unit. When budgeting, require providers to specify the billing basis and measure the total cost of a single video under identical inputs, rather than multiplying the list price by duration.

Divergences in Documentation Conventions and Items Going Offline This Week

Cross-verification from multiple sources confirms that provider official documentation is notably inconsistent on billing units. Different pages from the same vendor may use different price units, some documents do not clearly state whether failed requests are charged, and others mix "credits" with USD without giving a conversion ratio. Such information gaps force third-party price comparisons to rely on actual measurements rather than copying from documentation.

According to source information, some model versions or pricing tiers are about to go offline this week, involving older image models and specific video resolution tiers. Teams relying on fixed prices for long-term budgets need to watch for such retirement announcements to avoid sudden call failures or changes in cost structure.

Mixing credits and USD: Many platforms use internal credits as the billing unit but do not provide a fixed USD conversion ratio in their documentation, or the ratio varies by plan tier. This makes it impossible for third-party comparisons to directly convert credit prices into USD, leaving only reverse inference through top-up packages. Some documents even mix the two units on the same page without labeling them. Developers should require providers to supply a clear credits-to-USD exchange rate and validity rules during selection, to avoid hidden costs from prepaid credits expiring.

Practical Advice for Developers

Faced with this price landscape, selection strategy should shift from "comparing unit prices" to "comparing total cost of ownership." It is advisable to run small-scale A/B tests under real business loads, recording success rates, retry counts, and average cost per output, rather than looking only at list prices.

At the same time, write billing units into the technical selection document: specify the basis on which each image is billed, whether video is billed by result duration or resource duration, and how failed requests are handled. After scaling up calls, these details affect costs far more than unit price differences themselves.

For cost-sensitive products, consider using different provider combinations for images and video: on the image side, choose low-price tiers billed per successful generation; on the video side, prioritize stability and feature completeness. The price list is only the starting point—predictability is what production environments truly need.

One-line Summary

Price comparison of 19 image and video generation APIs: same 1K label with over 70x price gap, and hidden traps in video billing units.

Paragraph Summary

A price survey of image and video generation APIs from 19 providers shows that for a single image labeled 1K, quotes range from $0.003 to $0.22, a gap of over 70x; for 720p video, the per-second price ranges from $0.025 to $1.50, a span of 60x. The huge price differences stem not simply from pricing strategy but from inconsistent definitions of "1K," different billing granularity (per successful generation vs per API call), and conditions such as volume or prepaid requirements attached to low-price tiers. On the video side, divergences in the definition of "second" (result duration vs resource duration) and differences in frame rate and audio bundling make cross-vendor prices non-convertible. Inconsistent documentation conventions and opaque billing rules for failed requests further complicate comparisons. The article advises developers to shift from comparing unit prices to comparing total cost of ownership, measure success rates and retry costs under real loads, and watch for retirement announcements of old model versions and pricing tiers.