Hailuo AI API Pricing Guide for Production Video Workflows

Hailuo AI API pricing (also documented under MiniMax API pricing on the provider's developer platform) works through prepaid video-point packages rather than pay-as-you-go credit card billing. There is no per-request checkout: you buy a package, points get deducted from that balance when a generation succeeds, and the balance expires after 30 days. This guide gives you the current package prices, the point deduction table by model and resolution, a live TokenLab pricing comparison against Veo 3.1, PixVerse V6, and Seedance, and a straight answer on what a direct Hailuo API integration versus a TokenLab integration actually requires.

If you came here for a Hailuo API integration example or Hailuo API call code, jump to the "Calling Hailuo Directly vs. Through TokenLab" section below. There is no fabricated code in this article. Where MiniMax's public docs do not expose enough endpoint or payload detail to write a safe example, we say so and point you to the verification checklist instead of guessing.

Key Takeaways

---

Source Snapshot

Source	What it covers	Observed
MiniMax API video packages docs (platform.minimax.io/docs/guides/pricing-video)	Package prices, point deductions, RPM limits, failure/security-review non-deduction policy	2026-07-08
PixVerse Platform Docs (docs.platform.pixverse.ai/pricing-796039m0)	V6 credit rates per second by resolution and audio	2026-07-08
fal PixVerse V6 model page (fal.ai/pixverse-v6)	Per-second USD pricing and JS subscribe payload for fal-ai/pixverse/v6/text-to-video	2026-07-08
Google AI Gemini API pricing (ai.google.dev/gemini-api/docs/pricing)	Veo 3.1 Standard/Fast/Lite per-second pricing, Veo 3.0 deprecation date, audio-failure billing policy	2026-07-08
Runway API pricing docs (docs.dev.runwayml.com/guides/pricing)	Credit price and per-second credit rates for seedance2, gen4.5, veo3/veo3.1	2026-07-08
Kling AI Developer Platform pricing (kling.ai/dev/pricing)	Unit-based pricing; exact per-model cost needs provider-page verification at publish time	2026-07-08
TokenLab live model/pricing evidence	Direct per-request/per-second pricing for hailuo, veo-3, pixverse, seedance series	SSOT observed 2026-07-07, expires 2026-07-14

---

MiniMax Video Packages (Hailuo AI API Pricing)

MiniMax's developer platform sells four fixed video packages. There is no smaller starter tier and no monthly subscription that auto-renews at a lower commitment; every tier requires the full purchase up front.

Package	Price	Video Points	Validity	RPM Limit	Effective Cost / Point	Source	Observed
Standard	$1,000	3,760	1 month	20	~$0.266	MiniMax pricing docs	2026-07-08
Pro	$2,500	9,920	1 month	30	~$0.252	MiniMax pricing docs	2026-07-08
Scale	$4,500	18,900	1 month	40	~$0.238	MiniMax pricing docs	2026-07-08
Business	$6,000	26,780	1 month	50	~$0.224	MiniMax pricing docs	2026-07-08

The 1-month validity applies to every tier. MiniMax's published pricing page does not state whether unused points carry over if you purchase a new package before or after expiration. If your team plans monthly renewals, confirm the rollover policy with MiniMax support or your account rep before committing budget to a tier sized for peak rather than average usage.

If a 1-month expiring package does not fit your workflow, you do not have to buy one. Create a TokenLab API key and call hailuo-2.3-fast, hailuo-2.3, or hailuo-2.3-pro directly at the TokenLab model directory on a per-request basis, with no package purchase, no 30-day clock, and no RPM tier to plan around.

---

Point Deduction Rules by Model

Points are deducted only on a successful generation. The rate depends on model variant, resolution, and duration.

Model	Resolution / Duration	Points Deducted	Cost at Standard Package	Cost at Business Package	Source	Observed
hailuo-2.3-fast	768p / 6s	0.7	$0.186	$0.157	MiniMax pricing docs	2026-07-08
hailuo-2.3-fast	768p / 10s	1.1	$0.293	$0.246	MiniMax pricing docs	2026-07-08
hailuo-2.3-fast	1080p / 6s	1.3	$0.346	$0.291	MiniMax pricing docs	2026-07-08
hailuo-2.3 / hailuo-2.3-standard	768p / 6s	1.0	$0.266	$0.224	MiniMax pricing docs	2026-07-08
hailuo-2.3 / hailuo-2.3-standard	768p / 10s	2.0	$0.532	$0.448	MiniMax pricing docs	2026-07-08
hailuo-2.3 / hailuo-2.3-standard	1080p / 6s	2.0	$0.532	$0.448	MiniMax pricing docs	2026-07-08

MiniMax's docs state directly that "video generation failures or security-review videos do not result in a deduction" (MiniMax API video packages docs, observed 2026-07-08). That policy covers both outcomes explicitly; there is no separate disclosure needed for security-review rejections specifically, since the same sentence in the source covers both cases.

---

TokenLab Live Video Model Pricing

This table includes only models present in the current TokenLab live pricing evidence. Prices are per-request or per-second depending on how each provider bills, with no package purchase required.

Model	Provider	TokenLab Price	Unit	Source	Observed
hailuo-2.3-fast	MiniMax	$0.190000	per request	TokenLab live evidence	2026-07-07
hailuo-2.3-standard	MiniMax	$0.280000	per request	TokenLab live evidence	2026-07-07
hailuo-2.3	MiniMax	$0.280000	per request	TokenLab live evidence	2026-07-07
hailuo-2.3-pro	MiniMax	$0.490000	per request	TokenLab live evidence	2026-07-07
veo3.1-fast	Google	$0.080000	per second	TokenLab live evidence	2026-07-07
veo3.1	Google	$0.200000	per second	TokenLab live evidence	2026-07-07
pixverse-v6	PixVerse	$0.022059	per second	TokenLab live evidence	2026-07-07
seedance-2.0-fast	ByteDance	$5.441176	per output token	TokenLab live evidence	2026-07-07

Note: veo3.1 and veo3.1-fast on TokenLab are listed as flat per-second locks ($0.20/s and $0.08/s respectively), which differ from Google's own published Veo 3.1 tiered rates (Standard $0.40/s at 720p/1080p, Fast $0.10-0.12/s, Lite $0.05-0.08/s, all with audio). Confirm which resolution/audio tier the TokenLab rate maps to before budgeting a large render batch.

Compare the full landscape in TokenLab's best AI video models API 2026 roundup, or pair video with image generation using the best AI image models API 2026 guide. Browse live specs directly at TokenLab's video model category.

---

Other Video API Options Worth Checking

Runway and Kling also publish API pricing worth comparing before you lock into a MiniMax package:

• Runway sells API credits at $0.01 each. seedance2 runs 36 credits/s at 480p/720p and 40 credits/s at 1080p; gen4_turbo runs 5 credits/s; veo3.1 with audio runs 40 credits/s, without audio 20 credits/s (Runway API pricing docs, observed 2026-07-08).
• Kling's developer platform prices in "Units," with a snapshot showing 1 Unit at roughly $0.14 list price. Kling's exact per-model cost needs provider-page verification at publish time; do not quote a hard per-second Kling number without checking the live page first (Kling AI Developer Platform pricing, observed 2026-07-08).

---

What Actually Drives Production Costs

The advertised per-request or per-point rate is the floor, not the ceiling, of what a production pipeline spends. Two cost drivers matter in practice:

1. Iteration overhead. Text-to-video and image-to-video prompts often need more than one attempt to hit a usable result. There is no MiniMax-, PixVerse-, or Google-published figure for how many attempts a typical team needs per final clip, so treat any "3 to 5 variations" or "4x effective cost" number you see (including in earlier versions of this article) as an unsourced planning heuristic, not a measured benchmark. Run your own pilot batch and log attempts-to-acceptance before you build a budget model on it.
2. Resolution tiering mid-pipeline. Running a draft pass at 768p to validate composition, then a final render at 1080p, adds a second charge per approved clip. This is a real, calculable cost from the deduction tables above; unlike the iteration ratio, you can compute it exactly from your own resolution choices without guessing.

Storage and CDN delivery for finished video files are a separate line item from the generation API fee and should be budgeted independently.

---

Calling Hailuo Directly vs. Through TokenLab

There are two integration paths, and they are not interchangeable in terms of what evidence backs the code you'd write.

Direct MiniMax/Hailuo API. MiniMax's public pricing docs describe packages, points, and rate limits, but this evidence set does not include the exact request endpoint, authentication header format, or JSON request/response schema for a Hailuo generation call. Do not copy code from a blog post (including this one) that claims to show a direct hailuo-2.3-fast call unless it links to MiniMax's own API reference for the endpoint and payload. Before you integrate directly, verify:

• The exact base URL and versioned endpoint path in MiniMax's current API reference.
• The authentication scheme (API key header name, bearer token, or signed request).
• Whether generation is synchronous or job-based (most video APIs return a job ID and require polling or a webhook).
• The request schema for text-to-video versus image-to-video calls. Video generation is inherently multimodal when you submit a reference image alongside a prompt, and the exact image-to-video payload (image encoding, size limits, supported formats) must be verified in MiniMax's official docs before production use. Do not assume it mirrors PixVerse's or Veo's payload shape.
• How the API surfaces 4xx errors (bad prompt, invalid resolution), 429 (rate limit exceeded, relevant given the 20-50 RPM package caps), 503 (service unavailable), and timeouts on long-running jobs, plus what retry-with-backoff behavior MiniMax recommends.

Through TokenLab. TokenLab exposes hailuo-2.3-fast, hailuo-2.3, hailuo-2.3-standard, and hailuo-2.3-pro as billed line items in its live model directory, so you can route Hailuo calls through a single TokenLab API key alongside other providers instead of managing MiniMax's package-and-points billing separately. Create a TokenLab API key and browse current Hailuo rates to see current per-request pricing before you commit to a $1,000+ prepaid package.

Verified comparison example (PixVerse V6 via fal, not Hailuo). The one code sample below is backed by fal's own documented payload for PixVerse V6, included here as a working reference for how a per-second video API request is typically structured, not as a Hailuo integration:

// Verified against fal's PixVerse V6 docs (fal.ai/pixverse-v6, observed 2026-07-08)
// This is a PixVerse V6 call, not a Hailuo/MiniMax call.
try {
  const response = await fal.subscribe("fal-ai/pixverse/v6/text-to-video", {
    input: {
      prompt: "Cinematic shot of a futuristic city at sunset, 4k resolution",
      resolution: "720p",
      duration: 5,
      generate_audio_switch: false
    },
    logs: true
  });
  console.log("Generated video URL:", response.data.video.url);
} catch (err) {
  // Handle rate limiting, validation errors, and timeouts explicitly.
  if (err.status === 429) {
    // Back off and retry with exponential delay.
  } else if (err.status >= 500) {
    // Provider-side failure; retry a limited number of times, then fall back
    // to an alternate model or queue the job for later.
  } else if (err.name === "TimeoutError") {
    // Long-running video jobs may need polling instead of a single await.
  } else {
    // 4xx validation errors: log the prompt/params and do not retry blindly.
    console.error("PixVerse request failed:", err);
  }
}

If you need a Hailuo-equivalent code sample with a confirmed endpoint and payload, treat that as a follow-up task against MiniMax's live API reference or TokenLab's model documentation rather than something to copy from a pricing article.

For teams running several video and image providers at once, routing through an aggregator layer instead of maintaining each provider's SDK separately can reduce integration overhead. TokenLab's OpenRouter comparison covers this tradeoff for teams deciding between direct SDKs and a unified API layer.

---

Checklist Before You Commit

• Package expiration. Can your team consume the purchased points inside the 1-month window? Confirm rollover policy directly with MiniMax; it is not stated in the pricing docs reviewed here.
• Iteration ratio. Run a pilot batch (your own sample size, not a borrowed industry figure) to measure attempts per accepted clip.
• RPM ceiling. Standard tier caps at 20 RPM. Model your peak concurrent traffic against the tier's limit before buying up.
• Safety filter behavior. Test your prompt patterns for false-positive security-review blocks; confirmed non-deduction on rejection reduces but does not eliminate the operational cost of a blocked prompt (retries, manual review).
• Multimodal payload. If your pipeline uses image-to-video, verify the exact image input schema in MiniMax's official API reference before writing production code.
• Alternative billing. Compare per-second models (Veo 3.1, PixVerse V6, Seedance) against per-request Hailuo pricing for your specific clip lengths using the TokenLab table above.

---

Limitations

• No endpoint, authentication, or request/response schema evidence for a direct Hailuo/MiniMax API call was available in this evidence set; direct integration code is intentionally omitted.
• Rollover policy for unused points at package expiration is not documented in the reviewed MiniMax pricing page; verify with MiniMax directly.
• The commonly cited "3-5 variations per clip" and "4x effective cost" iteration figures are not sourced to any MiniMax, PixVerse, or Google publication; they are treated here as unverified heuristics, not benchmarked in this evidence set.
• TokenLab's flat per-second Veo 3.1 rates do not map cleanly to Google's tiered (Standard/Fast/Lite, resolution, audio) pricing; confirm which tier applies before large batch runs.
• Kling's exact per-model per-second cost requires provider-page verification at publish time and is not quoted as a hard number here.

---

FAQ

How much does Hailuo AI API pricing cost? Direct MiniMax packages start at $1,000 for 3,760 points ($0.266/point) up to $6,000 for 26,780 points ($0.224/point). A single hailuo-2.3-fast request at 768p/6s costs about $0.157-$0.186 depending on your package tier. Through TokenLab, hailuo-2.3-fast is $0.190000 per request with no package purchase required.

What's a Hailuo API integration example I can actually use? This article does not include a Hailuo-specific code sample because MiniMax's endpoint, auth, and payload details were not in the evidence reviewed. Use the checklist in "Calling Hailuo Directly vs. TokenLab" to verify those details against MiniMax's current API reference, or call the model through TokenLab's directory, where billing is per request with published rates.

What happens to unused points at the end of the 30-day validity period? Do they roll over if I buy another package? MiniMax's pricing docs state a 1-month validity per package but do not specify rollover behavior for unused points. Do not assume rollover; confirm directly with MiniMax before sizing a package around expected leftover balance.

Does MiniMax charge for failed generations or security-review rejections? No. MiniMax's own docs state that video generation failures and security-review videos do not result in a point deduction (MiniMax API video packages docs, observed 2026-07-08).

Where do I sign up or get API access? For direct MiniMax access, you purchase one of the four packages through MiniMax's developer platform. For per-request billing without a prepaid package, create a TokenLab API key and call Hailuo models alongside Veo 3.1, PixVerse V6, and other video APIs from one account.

How does TokenLab help compare Hailuo AI API pricing with alternatives? TokenLab maintains live per-request and per-second pricing for MiniMax, Google, PixVerse, and ByteDance video models in one directory, so you can normalize cost per clip across providers before choosing a model. See the TokenLab model directory for current rates.

---

Compare video models and create your TokenLab API key to start calling Hailuo, Veo 3.1, PixVerse V6, or Seedance on a per-request basis before committing to a prepaid MiniMax package.