Integrations

The AI models that power Scenes AI

Scenes AI generates through models built by Google, ByteDance, Kuaishou, MiniMax and Suno. You pick a model by what it is good at — visual quality, price, how long a single scene can run, or how many reference images it will accept — and Scenes AI handles the rest. Which models appear here comes from the same catalogue the product bills from, so this is always what you can actually use today.

Video

Animates each generated still into a clip. Every clip is image-to-video — the still comes first, which is exactly what lets your character and location references control what appears on screen.

  • Veo 3.1 Liteby Google

    Fastest and cheapest · ideal for iterating. Clips up to 8s, 2 aspect ratios, up to 3 reference images.

    Veo 3.1 Lite specs and pricing
  • Seedance 2.0 Miniby ByteDance

    Cheapest Seedance tier · multi-shot scenes up to 15s · capped at 720p. Clips up to 15s, 7 aspect ratios, up to 9 reference images.

    Seedance 2.0 Mini specs and pricing
  • Gemini Omni 1.1 Flashby Google

    First and last frame control · 7 reference slots · 1080p at the same price as 360p. Clips up to 10s, 2 aspect ratios, up to 7 reference images.

    Gemini Omni 1.1 Flash specs and pricing
  • Veo 3.1 Fastby Google

    Balanced quality and cost · strongest in-video audio. Clips up to 8s, 2 aspect ratios, up to 3 reference images.

    Veo 3.1 Fast specs and pricing
  • Wan 3.0by Alibaba

    Up to 30s in one generation · 1080p with native audio · 10 reference images. Clips up to 30s, 6 aspect ratios, up to 10 reference images.

    Wan 3.0 specs and pricing
  • Kling 3.0by Kuaishou

    Native multi-shot (up to 5 shots) · character binding with voice · up to 15s. Clips up to 15s, 3 aspect ratios, up to 12 reference images.

    Kling 3.0 specs and pricing
  • Seedance 2.0 Fastby ByteDance

    Fastest, cheapest Seedance tier · multi-shot scenes up to 15s · capped at 720p. Clips up to 15s, 7 aspect ratios, up to 9 reference images.

  • Seedance 2.0by ByteDance

    Best visual quality · multi-shot scenes up to 15s · 9 reference images. Clips up to 15s, 7 aspect ratios, up to 9 reference images.

    Seedance 2.0 specs and pricing
  • MiniMax H3by MiniMax

    Omni-reference (9 images + video + audio) · native audio · up to 15s. Clips up to 15s, 7 aspect ratios, up to 9 reference images.

    MiniMax H3 specs and pricing
  • HappyHorse 1.1by Alibaba

    Top-ranked motion and identity · 9 reference images · native audio with multilingual lip-sync · up to 15s. Clips up to 15s, 9 aspect ratios, up to 9 reference images.

    HappyHorse 1.1 specs and pricing
  • Seedance 2.5by ByteDance

    Longest single shot — up to 30s unbroken · 30 reference images · 1080p. Clips up to 30s, 7 aspect ratios, up to 30 reference images.

    Seedance 2.5 specs and pricing

Images

Generates the character sheets, the location plates and the still that opens every scene. Reference-image capacity matters more here than raw resolution, because references are how consistency is enforced.

Voice

Turns narration into speech. One engine takes a natural-language instruction for tone and emotion; the other covers the widest range of languages.

  • Gemini 3.1 Flash TTSby Google

    Thirty voices, with per-scene emotion and speaking rate.

  • ElevenLabs Multilingual v2by ElevenLabs

    Thirty voices, with per-scene emotion and speaking rate.

  • ElevenLabs Turbo 2.5by ElevenLabs

    Thirty voices, with per-scene emotion and speaking rate.

  • ElevenLabs — Your Voicesby ElevenLabs

    Thirty voices, with per-scene emotion and speaking rate.

Music

Writes an original track for your video rather than pulling one from a library, so the music follows the arc of your specific scenes.

You pick a model, not a vendor

You choose an engine by what it does — Seedance 2.0 for the best visual quality, Kling 3.0 for native multi-shot scenes and twelve reference images, Veo 3.1 Lite when you are still iterating and want it cheap. Everything behind that choice is infrastructure, and it can change without changing what you picked or what you are charged. See every model we support, the full price list, or how reference images keep characters consistent.