Hailuo 02: the complete guide

Hailuo-02 is a text-to-video and image-to-video model optimized for strong motion on characters, creatures and dynamic scenes at aggressive per-second pricing. It does not generate synchronized speech or lip-sync in the same pass — you trade native audio for one of the lowest API floors in the category.

• ▸Two API tiers on fal.ai: Standard outputs 768p; Pro outputs 1080p. Same prompt schema; different resolution and price.
• ▸Prompt optimizer (default on): the model rewrites your prompt before generation. Good for one-line briefs; turn off when you need exact wording preserved.
• ▸Duration is discrete: 6 seconds or 10 seconds. Pro 1080p is capped at 6 seconds per the API docs.
• ▸No native audio track — plan ElevenLabs, platform VO or royalty-free music in post.
• ▸Image-to-video uses a single start frame; no multi-reference stack like Seedance 2.0.

Official fal.ai billing for Hailuo-02:

• ▸A 6-second 1080p Pro clip at $0.48 is often cheaper than a single Veo 3.1 Fast 8-second clip at $0.96 — but you lose native audio and two seconds of runway.
• ▸Budget 2–3 attempts on complex architecture briefs; cheap per-second rates do not mean cheap per usable clip.
• ▸Third-party resellers (WaveSpeed, etc.) publish different per-clip bundles — verify against fal if that is your primary pipeline.

From our hero production: one 6-second Standard text-to-video generation for a creature close-up cost $0.27 on fal — the lowest line item in our homepage video stack.

Primary fal.ai endpoints:

curl -X POST https://queue.fal.run/fal-ai/minimax/hailuo-02/standard/text-to-video \
  -H "Authorization: Key $FAL_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Extreme close-up of a reptilian creature slowly turning its head, wet scales catching rim light, shallow depth of field.",
    "duration": "6",
    "prompt_optimizer": true
  }'

Queue-based async generation. Typical turnaround 1–4 minutes for a 6-second clip.

• ▸Character and creature work: describe skin texture, eye catchlight, head movement speed — Hailuo's docs and community reports favor tight subject framing over wide establishing shots.
• ▸With prompt_optimizer on, a short prompt often expands into a richer scene; test both on and off for brand-controlled wording.
• ▸Architecture: name structural details explicitly; budget extra attempts — our villa placeholder battles rate Pika highest failure rate, but Hailuo needs retries on complex geometry too.
• ▸No audio cues in prompt — they are ignored. Write a separate audio brief for post.
• ▸Image-to-video: use a high-contrast still with clear subject separation; motion amount is controlled via prompt verbs (slow pan, subtle breathing).

• ▸Wins: lowest documented per-second floor at 768p; strong motion on characters and creatures; simple API surface.
• ▸Loses: no native audio or lip-sync; shorter max duration than Kling or Seedance; less reference control than Seedance 2.0.
• ▸Loses to Veo on polish when the deliverable is a speech-led social ad — Veo Fast includes audio for ~$0.96 per 8 s.
• ▸Often beats Kling on raw cost per attempt, but Kling may need fewer retries on architecture flyovers.

Documentation verdict: Hailuo 02 is the budget workhorse — best when you need motion proof on a brief before spending Veo or Seedance credits. Watch for our budget-battle and creature-motion tests for ranked output quality.