AI video API prices per second (October 2026): 39 models compared
· Varosity
Live per-second prices for 39 AI video models (Veo 3.1, Kling, Seedance, Wan, Runway, HeyGen and more) at provider list price, with what a 5-second clip costs.
Every video model on Varosity is billed per second of video it produces, at the provider's list price with no markup. Pay with Varosity credits (1 credit = 1¢) or, where a model supports it, your own provider key, which the provider bills at the same rate. The table below is the live price list as of October 27, 2026, cheapest first. Prices follow the providers, so the models page is always current.
Text-to-video and image-to-video
| Model | Price per second | 5 s request costs | Sound | Lengths |
| LTX Video | $0.04 | $0.20 (5 s clip) | no | 5 s |
| Pika 2.2 | $0.04 | $0.20 (5 s clip) | no | 5/10 s |
| FLUX 3 Max (beta) | $0.0425 | $0.21 (5 s clip) | yes | 5–20 s |
| Veo 3.1 Lite | $0.05 | $0.20 (4 s clip) | yes | 4/6/8 s |
| Runway Gen 4 | $0.056 | $0.28 (5 s clip) | no | 5/10 s |
| WAN Video Effects | $0.06 | $0.30 (5 s clip) | no | 5/10 s |
| WAN 2.1 Text-to-Video | $0.06 | $0.30 (5 s clip) | no | 5 s |
| Vidu Q4 (beta) | $0.0665 | $0.33 (5 s clip) | yes | 3–16 s |
| Grok Imagine Video | $0.07 | $0.35 (5 s clip) | yes | 1–15 s |
| Luma Ray 2 | $0.08 | $0.40 (5 s clip) | no | 5/10 s |
| Hailuo 02 | $0.0833 | $0.50 (6 s clip) | no | 6 s |
| Hunyuan Video | $0.09 | $0.45 (5 s clip) | no | 5 s |
| HunyuanVideo 1.5 | $0.09 | $0.45 (5 s clip) | no | 5 s |
| Luma Ray 3.2 (fal) | $0.10 | $0.50 (5 s clip) | no | 5/10 s |
| HeyGen Cinematic Avatar (beta) | $0.10 | $0.50 (5 s clip) | yes | 4–15 s |
| Wan 2.6 | $0.10 | $0.50 (5 s clip) | no | 5/10/15 s |
| Wan 3.0 (beta) | $0.10 | $0.50 (5 s clip) | yes | 2–30 s |
| MiniMax H3 (beta) | $0.10 | $0.50 (5 s clip) | yes | 4–15 s |
| Runway Gen-4.5 (Replicate) | $0.12 | $0.60 (5 s clip) | no | 5/10 s |
| Seedance 2.0 Mini (beta) | $0.12 | $0.60 (5 s clip) | yes | 4–15 s |
| Seedance 1.0 Pro | $0.1215 | $0.61 (5 s clip) | no | 2–12 s |
| HappyHorse 1.0 | $0.14 | $0.70 (5 s clip) | yes | 5–15 s |
| Grok Imagine Video 1.5 (beta) | $0.14 | $0.70 (5 s clip) | yes | 1–15 s |
| Veo 3.1 Fast | $0.15 | $0.60 (4 s clip) | yes | 4/6/8 s |
| Kling 4.0 (beta) | $0.157 | $0.79 (5 s clip) | yes | 3–30 s |
| Kling 3.0 Pro | $0.168 | $0.84 (5 s clip) | yes | 5/10 s |
| Seedance 2.0 Fast | $0.2419 | $1.21 (5 s clip) | yes | 4–15 s |
| Seedance 2.0 | $0.3024 | $1.51 (5 s clip) | yes | 4–15 s |
| Kling 3.0 Pro (Replicate) | $0.336 | $1.68 (5 s clip) | yes | 5/10 s |
| Veo 3.1 (Quality) | $0.40 | $1.60 (4 s clip) | yes | 4/6/8 s |
"5 s request costs" is what a request for 5 seconds is charged, and how long the clip that comes back is. Models with fixed lengths fit your request to one of them: some take the nearest (a 5 s request on Veo 3.1 returns 4 s), some round up (6 s on Pika 2.2 returns 10 s). You pay for the seconds the clip actually runs, not for the length you asked for. GET https://varosity.ai/api/v1/video/estimate?modelId=…&durationSec=… (no key needed) shows the exact seconds and cost before you submit.
Avatars, lip-sync and motion transfer
| Model | Price per second | What it does |
| LivePortrait (Performance Transfer) | $0.05 | Moves a face in a photo to match a driving video (head, expression, lips) |
| LatentSync Lip-Sync | $0.05 | Lip-syncs an existing video to new audio |
| D-ID AI Presenter | $0.05 | A talking presenter from one photo |
| HeyGen Avatar (Digital Twin) | $0.08 | Your HeyGen digital twin reads a script |
| HeyGen Photo Avatar (Avatar IV) | $0.08 | Any photo becomes the speaker |
| HeyGen Video Translate | $0.10 | Dubs a video into another language with matching lips |
| HeyGen Video Agent (beta) | $0.12 | Writes the script, picks the avatar and scenes, and renders a finished video (length chosen by the agent) |
| OmniHuman 1.5 | $0.16 | A talking avatar from one photo and an audio track |
| Wan 2.2 Animate (Full-Body Transfer) | $0.16 | Moves a whole person in a photo to match a driving video |
Most of these run as long as the audio, script or driving video you give them, so the price per second is the useful number; HeyGen Video Agent picks its own length.
Which one to pick
- Cheapest drafts: LTX Video and Pika 2.2 at $0.04/s, silent. Good for testing a prompt before paying for a better model.
- Cheapest with sound: FLUX 3 Max (beta) at $0.0425/s, then Veo 3.1 Lite at $0.05/s (4, 6 or 8 s).
- Lip sync and talking heads: Veo 3.1 Fast ($0.15/s) has the best lip sync of the clip models; for a presenter from a photo, D-ID ($0.05/s) or HeyGen Photo Avatar ($0.08/s).
- Longer single clips: Wan 3.0 (beta) runs up to 30 s, Kling 4.0 (beta) 3–30 s.
- Top quality: Veo 3.1 Quality ($0.40/s) and Seedance 2.0 ($0.3024/s), both with sound.
Not sure? Ask your agent to call suggest_model with what you want; it ranks models from the same catalog.
Try one
From an agent connected over MCP (Claude, ChatGPT, Gemini, Cursor, Claude Code and others — setup):
Use Varosity to make a 6-second Veo 3.1 Lite video of rain on a café window at night, with sound.
Or call the API:
curl -X POST https://varosity.ai/api/v1/video/generate \
-H "Authorization: Bearer $VAROSITY_API_KEY" \
-H "Content-Type: application/json" \
-d '{"modelId":"veo-3.1-lite","prompt":"Rain running down a café window at night, warm lights blurred behind it","aspectRatio":"16:9","durationSec":6}'It returns a jobId; poll GET /api/v1/jobs/{jobId} until its status is succeeded (or failed). Every model's page lists its exact price, lengths, shapes and prompting tips. Most new accounts include a small starter credit, and top-ups start at $5.