AI Video Generation Pricing - April 2026
Normalized per-second pricing for Sora 2, Veo 3, Runway Gen-4, Kling 2.x, Luma Ray2, Seedance 2, and more - Kling and Haiper lead on cost.

TL;DR
- Cheapest API: Haiper Video 2.x at $0.033/sec (540p) - but the quality gap is real
- Best value at production quality: Kling 2.x Standard subscription, under $0.015/sec effective at 720p
- Most expensive: Veo 3.1 Standard at $0.40/sec at 720p and $0.75/sec for the original Veo 3
- Seedance 2 (ByteDance) just landed on fal.ai - $0.20/sec at 720p with audio included
Quick Verdict
For production workloads where quality matters, Kling 2.x on a subscription delivers the best cents-per-usable-second. The Premier plan ($64.99/month) works out to roughly $0.008-$0.015 per second of 720p video depending on the generation mode. Runway Gen-4 Turbo via API lands at $0.05/sec, which is reasonable for its output quality - the Unlimited plan ($76/month) adds a relaxed-queue "Explore Mode" that's worth it if you're generating at volume.
Veo 3 pricing from Google is the sticker shock of this comparison. At $0.75/sec for the original API model, a 10-second clip costs $7.50. The newer Veo 3.1 Fast drops to $0.10/sec at 720p, which is more usable, and Veo 3.1 Lite at $0.05/sec makes it competitive with Runway. Sora 2 at $0.10/sec is near the same tier as Veo 3.1 Fast - a 10-second 720p clip at $1.00.
For a broader quality comparison across these generators, see our video generation benchmarks leaderboard and the best AI video generators roundup.
Normalized Pricing Table
All prices are per second of output video. Where vendors use credits, I've converted at their published credit-to-dollar rates. Verified April 19, 2026.
| Model | Provider | 720p/sec | 1080p/sec | Per 5s Clip | Input Type | Audio Included |
|---|---|---|---|---|---|---|
| Haiper 2.x | Haiper | $0.033 | N/A | $0.17 | T2V, I2V | No |
| LTX-2 Fast | Lightricks | $0.04* | $0.04 | $0.20 | T2V, I2V | No |
| Kling 2.x Standard | Kling (sub) | ~$0.008-$0.015 | ~$0.015-$0.022 | $0.26 (PAYG) | T2V, I2V | Optional (+cost) |
| Runway Gen-4 Turbo | Runway | $0.05 | $0.05 | $0.25 | T2V, I2V | No |
| LTX-2.3 Fast | Lightricks | $0.06 | $0.06 | $0.30 | T2V, I2V, V2V | No |
| Veo 3.1 Lite | $0.05 | $0.08 | $0.25-$0.40 | T2V, I2V | Yes | |
| Runway Gen-4.5 | Runway | $0.25 | $0.25 | $1.25 | T2V, I2V | No |
| Sora 2 | OpenAI | $0.10 | N/A | $0.50 | T2V, I2V | No |
| Veo 3.1 Fast | $0.10 | $0.12 | $0.50-$0.60 | T2V, I2V | Yes | |
| Seedance 2 Fast | ByteDance | $0.24 | N/A | $1.21 | T2V, I2V, Ref2V | Yes |
| Hailuo 02 Standard | MiniMax | ~$0.045 | N/A | ~$0.23 | T2V, I2V | No |
| Seedance 2 Standard | ByteDance | $0.30 | N/A | $1.52 | T2V, I2V, Ref2V | Yes |
| Sora 2 Pro | OpenAI | $0.30 | $0.50 | $1.50-$2.50 | T2V, I2V | No |
| Veo 3.1 Standard | $0.40 | $0.40 | $2.00 | T2V, I2V | Yes | |
| Veo 3 | $0.75 | $0.75 | $3.75 | T2V | Yes |
*LTX-2 Fast has no distinct 720p pricing - it charges by output resolution starting at 1080p. At lower resolutions the cost scales down proportionally.
Subscription vs. API Pay-As-You-Go
Several vendors price subscriptions separately from their API. Kling and Luma in particular offer subscription credits that are meaningfully cheaper than their PAYG API rates.
| Service | Plan | Monthly Cost | Credits | Effective $/5s Clip | Commercial Use |
|---|---|---|---|---|---|
| Runway Standard | Sub | $12 | 625 credits | ~$0.19-$0.96 | Yes |
| Runway Pro | Sub | $28 | 2,250 credits | ~$0.19-$0.96 | Yes |
| Runway Unlimited | Sub | $76 | 2,250 + Explore Mode | ~$0.05+ (relaxed) | Yes |
| Kling Standard | Sub | $6.99 | 660 credits | ~$0.26 (5s std) | No |
| Kling Pro | Sub | $25.99 | 3,000 credits | ~$0.26 (5s std) | Yes |
| Kling Premier | Sub | $64.99 | 8,000 credits | ~$0.26 (5s std) | Yes |
| Pika Pro | Sub | $28 | 2,300 credits | ~$0.09-$0.35 | Yes |
| Luma Plus | Sub | $30 | 400 credits est. | ~$0.11-$0.38 | Yes |
Runway Pro at $28/month gives 2,250 credits. At $0.01/credit from their API, that's equivalent to $22.50 in API spend. You save roughly 20% versus PAYG. The real value is Unlimited tier with Explore Mode for high-volume generations at relaxed throughput.
Runway Gen-4's character consistency across scene cuts, one of the key differentiators for the Gen-4 line.
Source: runwayml.com
Per-Vendor Breakdown
OpenAI Sora 2
Sora 2 access is subscription-only through ChatGPT - there's no standalone free tier as of January 2026. ChatGPT Plus ($20/month) includes unlimited 480p access and limited 720p via a credit pool. ChatGPT Pro ($200/month) gives 10,000 credits monthly and supports up to 20-second clips at 1080p.
Via the API, Sora 2 standard runs $0.10/sec at 720p with 4-, 8-, or 12-second clip options. Sora 2 Pro hits $0.30/sec at 720p or $0.50/sec at 1024p for 25-second max duration. No native audio generation. Image-to-video and video-to-video cost the same rate as text-to-video.
Google Veo 3.x
The Veo family spans three price tiers. Veo 3 at $0.75/sec is the flagship - it produces native synchronized audio and dialogue. Veo 3.1 Fast at $0.10/sec (720p) or $0.12/sec (1080p) is the practical developer option. Veo 3.1 Lite, launched in 2026, undercuts at $0.05/sec (720p) and $0.08/sec (1080p) - making it comparable to LTX Studio's entry model.
Vertex AI prices Veo differently from the Gemini API. Developers should double-check which route their region uses. All tiers include audio. Google charges only for successful generations, so failed renders don't count against your bill.
Runway Gen-4 / Gen-4.5
Runway's API is purely credit-based at $0.01/credit. Gen-4 Turbo at 5 credits/sec works out to $0.05/sec - the same whether you're doing text-to-video or image-to-video. Gen-4.5 climbs to 25 credits/sec ($0.25/sec). The older Gen-4 Aleph hits 15 credits/sec ($0.15/sec).
Commercial use is permitted on Standard tier and above. Credits don't roll over, which is worth knowing for production budgeting. The Unlimited subscription at $76/month adds Explore Mode for unlimited relaxed-rate generation - a significant value if you're iterating on prompts or doing R&D rather than final production.
Kling 2.x (Kuaishou)
Kling uses a credit system priced at roughly $0.69-$1.06 per 100 credits depending on plan tier. Standard 5-second 720p video costs 10 credits ($0.07-$0.11); Pro mode with audio jumps to 60-100 credits per 5-second clip.
The free tier gives 66 credits/day - around 6 standard 5-second videos daily, permanently. That's the most generous ongoing free tier in this comparison. Consumer subscription plans range from $6.99/month (660 credits) to $180/month (26,000 credits). API access is available via the official Kling developer portal at klingai.com; third-party providers like PiAPI offer per-video rates from $0.26/5-second standard clip.
Commercial use requires Pro plan or above.
Luma Dream Machine (Ray2 / Ray3)
Luma's subscription credits don't transfer to the API - they run entirely separate billing systems. Web subscriptions start at $30/month (Plus). The API charges by pixel produced: $0.32/million pixels, though some providers quote it as $0.20 per video task.
Ray2 Flash costs 11 credits/sec and Ray2 Standard 32 credits/sec within the subscription system. Ray3.14 HDR at 1080p hits 768 credits for a premium clip. For a 5-second clip at Ray2 Standard, that's 160 credits - which on the Plus plan ($30/month) works out to roughly $0.38 per clip at effective credit rates.
ByteDance Seedance 2
Seedance 2 launched on fal.ai in April 2026, making its API publicly accessible outside China for the first time. The standard tier costs $0.3034/sec at 720p; the Fast tier is $0.2419/sec. A 10-second clip at standard quality runs ~$3.03 - putting it between Veo 3.1 Fast and Veo 3.1 Standard in price.
Audio is included at no extra cost in both tiers. Video-to-video (reference-based) generation applies a 0.6x multiplier on the input portion, making it cheaper than text-to-video at $0.18/sec. Official pricing in China is 46 CNY per million tokens (~$0.14/sec equivalent), so Western API access via fal.ai carries a roughly 2x markup.
Hailuo 02 (MiniMax)
MiniMax's Hailuo 02 Standard costs about $0.045/sec at 768p through third-party API providers, with a 6-second clip running around $0.25-$0.28. On Segmind's platform it's listed as $0.625 per full generation. The Pro tier improves temporal coherence and maxes out at 1080p. Consumer subscriptions start at $9.99/month for 1,000 credits.
Hailuo 02 supports both T2V and I2V. No native audio - you'll need a separate audio generation step for synchronized sound.
Pika 2.x
Pika moved to a subscription-first model with a small free tier (80 credits/month). Paid plans run $8/month (Standard, 700 credits) through $76/month (Fancy, 6,000 credits). Credit consumption varies significantly: the Turbo model uses 10-60 credits per video, the Pro model 20-80 credits.
Commercial use is permitted on all paid plans. Pika doesn't offer a standalone API - developers need to access it through third-party integrations. Image-to-video costs 15-18 credits per generation (roughly $0.15-$0.18 at $0.01/credit equivalent). No API PAYG option means this is subscription-only territory.
Haiper 2.x
Haiper's API is the cheapest pure-API option on this list at $0.033/sec (540p) and $0.05/sec (720p). The 540p tier is roughly 3x cheaper than Runway Gen-4 Turbo on a per-second basis. Quality is noticeably lower than the flagship models, but for high-volume low-fidelity workloads - social media drafts, storyboarding, iteration - the economics work.
Haiper supports T2V, I2V, and video extension. The 4K upscaling feature is listed as "coming soon." Consumer plans start at $8/month.
LTX Studio / LTX-2 (Lightricks)
LTX's API prices video by output duration at fixed rates per second. LTX-2 Fast at 1080p costs $0.04/sec; LTX-2.3 Pro at 1080p runs $0.08/sec. A special "Retake" video editing endpoint costs $0.10/sec on input video length, making it one of the few APIs with a native video-to-video editing mode.
LTX Studio subscriptions (the web interface) include Veo 2 access at Standard ($35/month) and Veo 2 plus Veo 3 at Pro ($125/month) - a good deal if you want Veo access without paying Google's per-second rate directly. The free plan offers 800 computing-seconds one-time.
Genmo Mochi
Genmo's Mochi 1 is open-source (Apache 2.0) and doesn't offer a commercial API service. Running it yourself on cloud infrastructure (like Modal with H100 GPUs) costs roughly $0.33 per generation. There's no managed API - you self-host or use it through community wrappers. For teams that want open-weight flexibility without the managed API overhead, it's worth considering. See the AI video generation capability guide for a technical breakdown of open vs. closed approaches.
AI-generated short film character showing what sub-$1-per-clip quality looks like in 2026.
Source: runwayml.com
Hidden Costs
Audio Is a Separate Line Item for Most Providers
Veo 3.x, Seedance 2, and Sora (via add-ons) include audio. Runway, Kling (base), Pika, LTX, Haiper, and Luma do not. Adding synchronized audio through a separate step - using ElevenLabs, Bark, or Kling's native audio add-on - costs $0.01-$0.05/sec on top of the video rate. For a 10-second clip at Runway Gen-4 Turbo rates, audio integration can double your effective cost.
Resolution Multipliers Are Non-Obvious
Most vendors charge by pixel area, not by a flat tier name. LTX scales linearly: 1080p at $0.04/sec doubles to $0.08/sec at 1440p and quadruples to $0.16/sec at 4K. Google Veo prices 4K at 2-3x the 1080p rate. If your pipeline ever outputs 4K for any reason, verify the pricing impact before routing production traffic.
Credit Expiration
Runway credits expire at month-end with no rollover. Kling credits don't roll over on Standard; they roll on Pro and above. Pika offers rollover credits starting at the Standard plan. Build expiration into your cost model if you have uneven generation volume.
Commercial Use Restrictions
Kling's Standard plan ($6.99/month) explicitly excludes commercial use. Pika permits commercial use on all paid plans. Runway permits it from Standard tier up. Luma and LTX permit commercial use on all paid tiers. Haiper permits commercial use on Pro ($24/month) and above. Failing to check this before shipping a client project is an avoidable legal exposure. Our best AI video editing tools guide covers commercial licensing in more depth for production pipelines.
At $0.75/sec, generating one minute of Veo 3 video through the API costs $45. A single YouTube Shorts series would run hundreds of dollars monthly.
Free Tier Comparison
| Provider | Free Offering | Amount | Limitations |
|---|---|---|---|
| Kling | Daily credits | 66 credits/day (~6 clips/day) | 5-second clips only, lower quality |
| Pika | Monthly credits | 80 credits/month | ~8 basic clips |
| LTX Studio | One-time allocation | 800 computing-seconds | Watermark, non-commercial |
| OpenAI / Sora | None (requires ChatGPT Plus $20+) | 0 standalone | - |
| Google Veo | Paid API tier only | 0 standalone | - |
| Runway | One-time credits | 125 credits | No Gen-4 Video |
| Haiper | Beta tier | ~10 daily creations | Lower resolution |
| ByteDance Seedance | Signup credits via Dreamina | 260 credits | China consumer platform |
Kling is the clear winner on ongoing free access. The 66 daily credits yield roughly 6 standard 5-second clips per day indefinitely - more than enough for experimentation and prototyping. No other provider comes close on sustained free volume.
Price History
Apr 2026 - ByteDance Seedance 2 launched on fal.ai, making the API available outside China at $0.24-$0.30/sec at 720p. Audio included. First serious competitor to Veo 3 on quality benchmarks.
Mar 2026 - Google released Veo 3.1 and Veo 3.1 Lite. Veo 3.1 Lite at $0.05/sec (720p) is 93% cheaper than the original Veo 3.
Jan 2026 - LTX Studio launched LTX-2.3 line with video editing ("Retake") at $0.10/sec. Runway added Gen-4.5 at $0.25/sec. Sora 2 Pro API went live at $0.30-$0.50/sec.
Dec 2025 - MiniMax launched Hailuo 02, improving 1080p quality notably. Kling hit version 2.6 with native audio support as an optional add-on.
Nov 2025 - Haiper launched API pricing at $0.033-$0.05/sec, undercutting most rivals on raw API cost. Luma launched Ray3 lineup.
Jul 2025 - Google Veo 3 API launched at $0.75/sec, the highest per-second rate of any major provider.
The market split into two tiers over the past year. Premium providers (Veo 3, Sora 2 Pro, Seedance 2) charge $0.30-$0.75/sec for native audio, physics fidelity, and consistent characters. Budget options (Haiper, LTX-2 Fast, Kling subscription) stay below $0.05/sec with acceptable output for many use cases.
FAQ
Which AI video API is cheapest?
Haiper 2.x API at $0.033/sec (540p) and $0.05/sec (720p). For subscription access, Kling's daily free credits deliver around 6 five-second clips per day at no cost. LTX-2 Fast at $0.04/sec (1080p) is the cheapest option with 1080p output.
What is the best value for production video generation?
Kling 2.x Premier subscription ($64.99/month, 8,000 credits) for subscription workflows, or Runway Gen-4 Turbo API ($0.05/sec) for direct API integration. Both deliver quality suitable for most commercial productions at under $0.10/sec effective.
Does Veo 3 cost $0.75 per second?
The original Veo 3 model does. Veo 3.1 Fast costs $0.10/sec (720p) and $0.12/sec (1080p). Veo 3.1 Lite runs $0.05/sec (720p). Google clearly dropped per-second rates with the 3.1 updates.
Can I use AI-created video commercially?
Most paid plans allow it, but check per-tier restrictions. Kling Standard ($6.99/month) excludes commercial use. Haiper requires the Pro plan ($24/month). Runway, Luma, LTX, and Pika allow commercial use on all paid tiers. Never assume - verify before shipping client work.
What's the difference between T2V, I2V, and V2V pricing?
Text-to-video (T2V) is the base rate at most providers. Image-to-video (I2V) costs the same at Runway, Sora, Veo, and Haiper. Video-to-video (V2V/reference) at Seedance 2 applies a 0.6x multiplier on the input portion. LTX's Retake endpoint prices by input video length at $0.10/sec, regardless of the standard generation rate.
Is Genmo Mochi available via API?
Not as a managed service. Mochi 1 is open-source under Apache 2.0 and runs on self-hosted infrastructure or community cloud deployments. Running it on a H100 via Modal costs roughly $0.33 per generation but requires setup effort. There's no official managed API endpoint.
Sources:
✓ Last verified April 19, 2026
