Hailuo AI — MiniMax’s Physics-Leading Text/Image-to-Video Generator (Hailuo 2.3 / 02 / H3, 6–15s Clips, Cheapest $/Clip vs Sora/Runway/Kling, Daily Free Credits)

Hailuo AI Consumer video brand of MiniMax Group (Beijing/Shanghai, founded 2021–22 by Yan Junjie + Yang Bin + Zhou Yucong, Hong Kong IPO Jan 9 2026, ~ $4B valuation, backers Alibaba/Tencent). Flagship models Hailuo 2.3 / 02 / H3 (Jul 2026). No acquisition, no sunset. Category `ai-productivity-data`, `pipeline_stage=none`. This is the **AI scene-video generation** lane sibling to your confirmed **PixVerse** card — Hailuo leans physics-accurate T2V/I2V + character consistency + cheapest$ /clip in the Sora/Runway/Kling tier; PixVerse leans creator-short-form effects. Not an avatar-presenter tool (≠ Synthesia/DeepBrain/Elai/Vidnoz), not TTS (≠ Murf/Speechify).

Official site: https://hailuoai.video (intl) / https://hailuoai.com (CN)

API: https://platform.minimax.io (MiniMax Open Platform, base api.minimax.io/v1)

Snapshot

Field	Value
Status	Active (Hailuo 2.3 live 2026, H3 Jul 31 2026, free daily credits + paid tiers + API live Aug 2026)
Origin	MiniMax Group, China 2021–22, HK IPO Jan 2026; Hailuo = consumer video face of MiniMax
Lane	AI scene-video generation (T2V/I2V/subject-ref/first-last-frame), not avatar-presenter, not TTS
Models	Hailuo 2.3 (physics flagship, WorldModelBench #1), 2.3-Fast (50% cheaper), Hailuo 02 (cinematic NCR 1080p), H3 / MiniMax H3 (Jul 2026, omni-modal, native 2K, 15s, stereo audio, Omni-Reference)
Output	720p(free) → 1080p → 2K(H3), 6s standard / 10s paid / 15s H3, 24–30fps, 16:9/9:16/1:1
Standout	Best-in-class physics sim (fluid/fabric/hair), character consistency across clips, ~ $0.30 / 10s 1080p clip (vs Runway$ 1.20, Sora $3.50), daily free 50–80 credits (~2–3 watermarked 768p clips/day)
Compliance	Data may be processed in China (GDPR gray zone for EU/US sensitive data); no SOC2/HIPAA like Murf/Synthesia; ongoing Disney/Universal/WB copyright suit (2025, dismissed-motion pending)
Category	`ai-productivity-data` / AI-scene-video sublane (sibling to PixVerse)

Features (table)

Layer	Capability	Notes
Text-to-Video	Prompt (≤2000 char) → 6–10s clip, camera commands (push/pan/tilt/track/static/zoom/shake), Director Mode, AnyAngle	Physics-leading motion
Image-to-Video	Single ref image → animated clip preserving comp/lighting; first-last-frame control
Subject Reference	Lock character/product face across generations (consistency leader w/ Kling)
Hailuo 2.3 / 02	2.3 = physics/micro-expression/anime/ink-wash; 02 = cinematic 1080p NCR	Fast variant 768p 6s
H3 (Jul 2026)	Omni-Reference (9 img + 3 vid + 3 audio), native 2K 15s, stereo audio same-pass, instruction editing (AA #1 video edit)	Weights partially open (MiniMax Community License)
Audio	Bundled TTS + voice clone in MiniMax Audio (separate sub-product), H3 emits native SFX/dialogue; no music sync in 2.3	Pair w/ ElevenLabs/Murf for clean VO
Media Agent	Multi-modal idea→auto-selects model per step (text/image/video/audio)
Prompt tools	Built-in prompt enhancer + DeepSeek integration for co-writing
No	No timeline editor, no avatar presenter, no LMS/SCORM, no real-time, 6–10s cap (2.3), text-in-video unreliable, hand/finger close-ups weak	Stitch externally

Pricing (2026-09, USD — credit-based, monthly reset no rollover)

Tier (Hailuo web)	Price	Credits / clips
Free	$0 (no card)	~500 signup + 50–80 daily (~2–3 × 768p 6s clips/day), watermark, no commercial rights, 720p, lowest queue
Standard	$14.99/mo (promo$ 7.99)	1,000 credits/mo (~40 × 6s 768p or ~12 × 1080p), no watermark, commercial, 1080p, 6s
Pro	$54.99/mo (promo$ 24.99)	4,500 credits (~180 clips), 10s 1080p, 2 parallel, priority queue
Master	$94.99/mo (promo$ 63.99)	10,000 credits (~400 clips)
Ultra	$124.99/mo	12,000 credits, free Hailuo 01 runs
Max	$199.99/mo	20,000 credits (~900 clips), all models
MiniMax API	Pay-as-you-go	~ $0.18/sec @768p, ~$ 0.26/sec @2K; M-series LLM $0.30/1M tok; Token Plan$ 10–50/mo dev sub
H3 API	2 concurrent free / 15 paid; no free clip quota documented

starting_price=$0; first paid ** $7.99/mo** (Standard promo). Per-clip ~$ 0.30/10s1080p — cheapest in tier.

Reviews (2026)

Source	Signal
AISO Tools 4.5 video / 4.3 motion / 4.2 value, AIScopeLab 4.6, ToolRadar mid-high	Praised: physics realism (water/fabric/hair), character consistency, anime/ink-wash styling, 30–90s gen, price/quality ratio vs Sora/Runway, generous daily free. Criticized: complex prompt adherence inconsistent, 6–10s cap (vs Runway 16s/Sora 60s), text-in-video unreliable, hand/finger close-ups, aggressive moderation, failed gens still burn credits, Trustpilot 1.4/5 billing/cancel complaints, data-in-China privacy, Disney/Universal suit pending
Hailuo vs Sora/Runway/Kling	Sora = length/polish, Runway = editing tools, Kling = human subjects; Hailuo = physics/stylized/character + cheapest
Aggregate	Pick for social shorts, anime/illustration, product-physics demos, character-driven series, rapid iteration on budget. Skip for client-brand face (Synthesia/DeepBrain), long-form (>15s), regulated/PII content (data-in-China), or teams needing timeline/DAW

Access / Integration (table)

Method	Supported
Web	hailuoai.video app, CN hailuoai.com
Mobile	iOS/Android Hailuo app
Import	Text prompt, image (JPG/PNG ≤10MB), first/last frame, subject ref, audio ref (H3)
Export	MP4 (watermark free / clean paid / zero-watermark via API)
API	REST `api.minimax.io/v1/video_generation` + query poll, Bearer key, webhook callback, Python/JS SDKs, OpenAI-compatible chat completions for M-series
Bundled	MiniMax Audio (TTS/clone), Music gen, Image gen, M2.7 LLM — all under one Token Plan key
No	No LMS/SCORM, no MCP, no timeline, no avatar presenter mode

Best For / Not For

Best For	Not For
Short-form social creators (TikTok/Reels/Shorts) at volume	Long-form (>15s) narrative (→ Runway/Sora)
Anime/illustration/ink-wash/Game CG styling	Client brand-face spokesperson (→ Synthesia/DeepBrain)
Product demos w/ liquid/fabric physics	Regulated/PII/health content (data-in-China, no HIPAA)
Character-driven series needing face consistency	Teams wanting timeline+broll compositing (→ Veed/Captions)
Devs wiring T2V/I2V into pipelines (cheap API)	Studio voiceover need (→ Murf/Speechify)
Rapid prompt iteration on $0 budget (daily free)	Wrong lane: Dev/eval/obs

Competitors (lane map)

Lane	Tool	Hailuo edge / gap
AI scene-video	Sora (OpenAI)	Sora longer (60s)/polish; Hailuo 3× cheaper, physics-leading, free daily
AI scene-video	Runway	Runway mature editing/controls; Hailuo cheaper/clip, physics+character consistency
AI scene-video	Kling AI	Kling stronger human subjects; Hailuo stronger physics/stylized/anime
AI scene-video	PixVerse (in batch)	PixVerse creator-effects/short-form social; Hailuo physics+consistency+API breadth
AI scene-video	Pika, Luma Dream Machine, Veo	Pika MCP/agent-y, Luma motion, Veo photoreal; Hailuo price/physics
Avatar video	Synthesia/DeepBrain/Elai/Vidnoz	Those = presenter/avatar; Hailuo = scene/no-talent
In-list overlap	PixVerse (active, scene-video sibling), all avatar/audio cards	Scene-video sublane only