AI Video Generation API Pricing 2026: Runway, Veo, MiniMax & Luma
Compare current developer pricing for Runway Dev, Google Veo 3.1, MiniMax H3 and Luma Ray 3.2. The focus is programmatic video generation economics—not consumer subscription plans and not a model-quality ranking.
Evidence basis: official developer pricing and documentation checked September 22, 2026. Cost examples use published rates and stated assumptions; they are not hands-on quality scores. Methodology · Sources & corrections
Fast answer: these APIs do not all meter video the same way. Runway converts developer credits to dollars, Google and MiniMax publish per-second video rates, and Luma prices Ray 3.2 by request type, resolution and duration. For budgeting, normalize the exact model and output you intend to ship rather than comparing a consumer plan's monthly sticker price.
Video generation API pricing at a glance
| API | Published billing model | Example public rate | 10-second example |
|---|---|---|---|
| Runway Dev | $0.01 per developer credit; many video models bill per output second | Gen-4.5: 12 credits/sec = $0.12/sec | $1.20 for 10s of Gen-4.5 |
| Google Gemini API / Veo 3.1 | USD per generated second | Veo 3.1 Standard: $0.40/sec at 720p or 1080p | $4.00 for 10s at the standard 720p/1080p rate |
| MiniMax Open Platform / H3 | USD per generated second, with separate input-material rules | H3: $0.08/sec at 768p; $0.13/sec at 2K | $0.80 at 768p or $1.30 at 2K |
| Luma Agents API / Ray 3.2 | Per generation by request type, resolution, dynamic range and duration | Standard 720p: $0.30/5s or $0.90/10s | $0.90 for a 10s standard 720p generation |
Important: this is a cost normalization, not an apples-to-apples quality benchmark. The compared examples differ in model family, resolution and feature set. A cheaper request can still cost more per approved clip if you need more retries, higher resolution, audio, edits or reference inputs.
Runway Dev pricing
Runway Dev sells developer credits at $0.01 per credit. Its current pricing page lists Gen-4.5 at 12 credits per second, equivalent to $0.12 per generated second, and Gen-4 Turbo at 5 credits per second, equivalent to $0.05 per second. Runway also exposes other video models—including Veo variants—through the same developer surface, each with its own credit rate.
That makes Runway relatively easy to model in a cost calculator: duration × model credit rate × $0.01, then add any documented surcharges or input costs. Runway's developer tiers also govern concurrency, daily generations and monthly spend. The company documents automatic tier increases based on spend, while queued tasks can be throttled when concurrency is full.
Runway Dev pricing → · Runway API usage tiers →
Google Veo 3.1 API pricing
Google prices Veo 3.1 directly per generated second. The current Gemini API pricing page lists Veo 3.1 Standard at $0.40/sec for 720p and 1080p and $0.60/sec for 4K. Veo 3.1 Fast is listed at $0.10/sec for 720p, $0.12/sec for 1080p and $0.30/sec for 4K. Veo 3.1 Lite is listed at $0.05/sec for 720p and $0.08/sec for 1080p, with no 4K output.
Google's Veo documentation identifies the 3.1, Fast and Lite preview model IDs and supports text or image input with video output that includes audio. Rate limits depend on the project, usage tier and model and are visible in AI Studio rather than being one universal number for every account.
Google also notes that when an audio-processing issue prevents a Veo video from being generated, you are charged only if the video is successfully generated.
Gemini API pricing → · Veo 3.1 API guide → · Gemini rate limits →
MiniMax H3 API pricing
MiniMax's Open Platform currently lists H3 at $0.08 per second for 768p and $0.13 per second for 2K. That means a 10-second output is $0.80 at 768p or $1.30 at 2K before any chargeable input materials.
H3's input-material billing matters for production economics. MiniMax lists audio input as free, the first five input images as free and additional images at $0.04 each. Video input is billed by input duration and output resolution. The platform also distinguishes the standard pay-as-you-go API key from subscription/token-plan access, so do not assume quotas from one surface automatically transfer to the other.
MiniMax API pricing → · MiniMax paid-services terms →
Luma Ray 3.2 API pricing
Luma's current Agents API prices Ray 3.2 per generation, not as one universal linear per-second rate. Standard video generation at 720p is listed at $0.30 for 5 seconds and $0.90 for 10 seconds; 1080p is $1.20 for 5 seconds and $3.60 for 10 seconds. HDR and HDR + EXR are separate 5-second tiers, while video edits, extend and reframe use their own billing tables.
This matters when normalizing costs: dividing a 5-second Luma request by five and extrapolating to ten seconds would be wrong because the published 10-second total is not simply twice the 5-second total. Use the provider's exact request type and duration table.
Luma documents refunds for several failed-generation codes such as content moderation, generation failure and output-not-found, while a budget-exhausted failure can have a partial charge. Pay-as-you-go has no minimum spend; Provisioned Throughput is a separate capacity product for larger production workloads.
Luma Agents API pricing → · Luma billing FAQ →
What changes the real cost per usable video?
- Retries: the nominal generation price is not the same as the cost per approved clip. Track rerolls and rejected outputs.
- Resolution and duration: Google, MiniMax and Luma publish materially different rates by output tier; Luma also changes price by request duration.
- Audio: Veo variants and Runway-routed Veo models can have distinct audio economics; check the exact model route.
- Reference media: MiniMax can charge for extra images or input video; other APIs also apply model-specific reference/input rules.
- Editing and extension: Luma separates generation, edit, extend and reframe. Runway also prices models and recipes independently.
- Throughput: concurrency, daily quotas and spend caps can constrain production even when per-clip pricing looks attractive.
- Failure billing: refund rules are provider- and failure-specific. Model expected failure/retry behavior before committing volume.
Which billing model is easiest to forecast?
Per-second billing is simplest when your outputs have predictable durations: Google and MiniMax expose this directly, while Runway's credit system converts cleanly to dollars because one developer credit is $0.01. Per-generation tables such as Luma's require you to model exact duration, resolution and operation type.
For production planning, build a spreadsheet or internal meter that stores provider, model, request type, resolution, requested duration, successful duration, retries, total API cost and approved output. The most useful metric is usually cost per approved clip, not cost per raw generation.
Do not mix developer API pricing with consumer plans
Consumer subscriptions and developer APIs can be separate commercial systems. Runway Dev sells developer credits directly; MiniMax explicitly distinguishes standard pay-as-you-go API keys from Token Plan subscription keys. Google bills Gemini API usage through its developer billing system, and Luma separates API pay-as-you-go/Provisioned Throughput from consumer creative plans. Treat every provider's API balance and consumer subscription as independent unless its documentation says otherwise.
Voice and short-form APIs are separate cost layers
If your workflow also needs TTS, dubbing or short-form post-processing, budget those as secondary layers rather than mixing them into the generative-video API ranking above. AI Video Signal has verified tracked routes for some of these providers:
For detailed voice/API pricing, use the dedicated ElevenLabs API pricing and Murf API pricing guides. For short-form automation, see Submagic API pricing.
Official sources checked
- Runway Dev — API Pricing & Costs
- Runway Dev — API Usage Tiers & Limits
- Google Gemini API — Pricing
- Google — Generate videos with Veo 3.1
- MiniMax Open Platform — API Pricing
- Luma Agents API — Pricing
- Luma Agents API — Billing FAQ
Related buyer checks before deployment
After choosing the API meter, verify commercial-use rights, model the retry budget with the AI Video Cost Calculator, compare current consumer-plan economics separately in the AI Video Pricing Tracker, and use Best AI Video Generators when you need workflow coverage rather than API billing.