AI生成 into the corner of the output under China’s labelling rules — NovAI cannot switch it off. Video output varies by model: in our 4 August 2026 tests Seedance 2.0 and CogVideoX-Flash showed no visible label, while Hunyuan Video 1.5 is labelled. Text and chat models are not affected. Which models, and what it rules out →We run NovAI, one of the providers compared below — first-party post, and every number is from a published pricing page (July 2026). Prices change; always re-check before committing volume.
Providers quote per-clip, per-second, or per-million-tokens. The only fair way to compare is normalized $/second at the same resolution. That is what this table does.
| Provider | Model | $/second (1080p) | 5s clip |
|---|---|---|---|
| fal.ai | Seedance 2.0 | ≈ $0.685 | ≈ $3.43 |
| poyo.ai | Seedance 2.0 | ≈ $0.45 | ≈ $2.25 |
| NovAI | doubao-seedance-2.0 | $0.329 | $1.65 |
| NovAI | doubao-seedance-2.0-fast | $0.264 | $1.32 |
Sources: fal.ai and poyo.ai published Seedance 2.0 pricing (poyo's own comparison page lists the fal numbers we cite); NovAI prices from our live catalog.
| Provider / model | Published price |
|---|---|
| Kling developer portal (per-second billing) | from ≈ $0.112/s; with native audio ≈ $0.14/s |
| fal.ai — Kling 2.1 Standard | $0.25 per 5s clip (+$0.05/extra second) |
| fal.ai — Kling 2.1 Pro | $0.45 per 5s clip |
| fal.ai — Kling 2.6 Pro (with audio) | $0.168/s |
NovAI — hy-video-1.5 (Tencent Hunyuan) | $0.21 per generation |
NovAI — cogvideox-flash | $0 per generation (free tier) |
| Resolution | doubao-seedance-2.0 | doubao-seedance-2.0-fast |
|---|---|---|
| 480p | $0.067/s | $0.054/s |
| 720p | $0.163/s | $0.131/s |
| 1080p | $0.329/s | $0.264/s |
| 4K | $1.00/s | $0.802/s |
Scenario: 100 × 5-second 1080p clips per month (short-form ads, product videos, storyboard iterations):
| Route | Monthly cost |
|---|---|
| fal.ai Seedance 2.0 | ≈ $342.50 |
| poyo.ai Seedance 2.0 | ≈ $225.00 |
NovAI doubao-seedance-2.0 | $164.50 |
NovAI doubao-seedance-2.0-fast | $132.00 |
A practical volume workflow: iterate prompts on the free cogvideox-flash tier or 480p drafts ($0.054/s), then render finals at 1080p. That routinely halves the bill again.
No reseller chain. We buy directly from the upstream provider (Volcano Engine / ByteDance) through a Chinese entity and publish the passthrough rate. Aggregators that source through intermediaries stack margins on top.
import requests, time
BASE = "https://aiapi-pro.com/v1"
KEY = {"Authorization": "Bearer sk-novai-xxx"}
task = requests.post(f"{BASE}/video/generations", headers=KEY, json={
"model": "doubao-seedance-2.0",
"prompt": "cinematic drone shot over a coastal highway at golden hour",
"resolution": "1080p",
"duration": 5,
}).json()
while True:
r = requests.get(f"{BASE}/video/generations/{task['id']}", headers=KEY).json()
if r.get("status") not in ("queued", "processing"):
break
time.sleep(5)
print(r)Swap model for cogvideox-flash to test the whole pipeline at $0 first — guide: Free AI Video Generation API in 2026.
Test free, scale at $0.329/s for 1080p Seedance 2.0.
NovAI prices are our live catalog rates as of July 2026. Competitor numbers are their published list prices on the same date — they can change, so verify before large commitments.
Yes: free cogview-3-flash, plus Seedream 5.0 ($0.027/image) and Hunyuan image models. See the free image API guide.
Video generation is async on every provider. We expose standard task polling and bill only when the upstream accepts the task.