AI Video Generator: MiniMax H3 & Seedance | Mazee.ai
Skip to content
Mazee.ai
ExploreModelsPricingResources
Sign inStart creating
Menu
ExploreModelsPricingResourcesSign in
  1. Home
  2. AI video creation
简体中文

AI videos with sound

Start with MiniMax H3 for member video creation, explore FastH3 for iteration, or choose Seedance for credit-based shots. Watch original clips before choosing a workflow.

Open member video studioWatch original video examples
MiniMax H3 · 7:4 · 768p

Video requires an eligible plan. Availability, supported region and content review apply; check your studio before submitting.

What you can create here today

MiniMax H3 and FastH3

Creator and Pro include designated H3 and FastH3 generation, subject to availability and fair queuing. There is no free video allowance. Credit packs alone do not grant video access.

New H3 tasks use Draft resolution: 5 or 8 seconds. Default landscape size is below; older samples can have higher recorded resolutions.

1024 × 576 · 5 / 8s · 24 fps

FastH3 is a separate model for faster iteration, with 720p and first/last-frame inputs here. It trades some motion and audio fidelity for speed. It is not a guaranteed quality or delivery-time upgrade.

Open member video studio

Seedance 2.5 and Seedance 2.0 Fast

Create 720p clips lasting 4–10 seconds from text, first/last frames or image/video references. Seedance 2.5 also supports source-video editing. Up to five images and one video are supported, with lower limits in some modes.

Seedance 2.5 · 720p · 5s · 315 credits estimated

Seedance uses creation credits and requires an active video-enabled plan. Credit packs add credits without changing your plan. Review the estimate again if you change the duration or model.

Open Seedance 2.5
Compare plans and creditsModel license and service terms

Watch original video examples

Play with sound. Each AI clip keeps its original prompt and recorded settings; results vary, and older samples may use different settings from today’s studio.

Hear the first sip

Recorded sample: MiniMax H3 · 768p · 7:4 · 5s

A product opening shot: the original prompt gives the hand one action and names the tab click and fizz separately. When reviewing, check the fingers, can shape and timing of the sound.

Read the original prompt
Single continuous five-second macro shot. One completely plain tangerine-orange metal soda can stands on wet charcoal slate. Its smooth orange body has absolutely no printing, letters, symbols, label or logo. Close framing shows the silver lid and orange shoulder against a featureless dark background. One adult hand slowly lifts the silver pull tab, opens the lid and withdraws. A small controlled fizz escapes; condensation trembles. The camera stays close, steady, warm rim light and cool fill. Keep one cylindrical can, correct hand anatomy and a blank orange surface throughout. Synchronized metallic tab click, brief sparkling fizz, quiet studio room tone. No speech, music, text, captions or watermark.
Inspect this sample

Then who has the key?

Recorded sample: Seedance 2.5 · 720p · 16:9 · 5s

A short suspense reveal: the matching keys give the scene one visual question. Check the number and shape of the keys, hand continuity and whether the reveal is readable before the clip ends.

Read the original prompt
One intimate medium shot, about five seconds, original suspense drama. An adult woman outside a quiet apartment holds a brass key up at chest height. A second key turns audibly in the lock from the unseen inside. She freezes, looks at her own key, then says softly in English, 'Then who has the key?' Slow restrained push-in, amber hallway light, natural facial expression and hand anatomy, consistent key and door. Sound: synchronized lock click, clear quiet dialogue, room tone. No music, captions or watermark.
Inspect this sample

Coffee shop dialogue

Recorded sample: FastH3 · 720p · 16:9 · 5s

A dialogue study: listen to the complete clip, then check speech timing, mouth movement and whether the camera keeps the speakers readable. This is a FastH3 sample, not an H3 quality comparison.

Read the original prompt
A locked medium close-up of one adult barista facing the camera in a quiet cafe, warm natural light, mouth clearly visible. The barista looks into the lens and speaks one complete English sentence in a calm, clear, natural American voice: "The coffee is ready whenever you are." Begin speaking around 0.5 seconds, finish by 4 seconds, then smile with lips closed. Natural lip sync, steady speaking pace, clear diction. Clean foreground dialogue. No music, subtitles, extra words, overlapping voices or loud background sounds.
Inspect this sample

Choose a workflow for the shot

Start with the input you already have. For a text-led shot with sound, inspect H3. For first/last frames, inspect FastH3 or Seedance. To change an existing video, use Seedance 2.5’s source-video mode. Compare cost and the exact input controls before committing to a model.

Practical video guides

Write a MiniMax H3 Prompt for a Shot with SoundSeedance First and Last Frames, References and Video EditsMiniMax H3 vs FastH3 for Your Next ShotKeep AI Video Dialogue and Characters Consistent Across Shots

From a video idea to a downloadable clip

  1. Choose the model for the shot

    Select H3 or FastH3 for designated member generation, or Seedance for a credit-based shot. Compare the original clips above before choosing.

  2. Write one readable action

    Describe the subject, one main action, camera movement and sound. With a first frame, describe what should move rather than only repeating its appearance.

  3. Check, generate and review

    Confirm the model, duration, resolution and cost. Follow the original task, then play the result with sound and save it from your private library.

Questions before you generate

Which model fits my input?

H3 is a starting point for text-led audio-video shots. FastH3 and Seedance support first/last frames here; Seedance 2.5 also edits a source video. Use the model pages for original samples and current limits.

Is video included in free image creation?

No. Free image access is separate. Video requires an eligible plan; designated H3/FastH3 usage and credit-based Seedance have different billing. Compare plans before opening a new task.

Does a five-second clip generate in five seconds?

No. Clip duration is the playback length. Generation also includes preparation, model loading, processing and saving; the task shows the recorded state.

Explore related workflows

Create videos with MiniMax H3Create videos with SeedanceAI product imagesReference image help

Model documentation

Capabilities checked against the shipped website settings and official sources on November 18, 2026.

  • MiniMax H3 model card
  • FastH3 workflow guide
  • Seedance 2.5 provider API
Mazee.ai

Creative tools

  • Image
  • Video
  • Models
  • Pricing & credits
  • Creation guides

Use cases

  • AI Ecommerce
  • AI Design
  • AI Game
  • AI Film
  • AI Art
  • AI Short Drama

Featured models

  • MiniMax H3
  • Seedance
  • Seedream 5.0 Flash
  • GPT Image 2.5 Flare
  • Nano Banana 2
  • Seedance 2.5
  • Kling 3.0

Support and policies

  • Help
  • Support requests
  • Policies & contact
  • Acceptable use policy
  • Privacy notice
  • Service terms notice
  • Storage notice
  • Email support: support@mazee.ai

Mazee.ai is an independent interface for third-party AI models. Model names describe integrations; we are not affiliated with or endorsed by their creators.