SoraOpenAI video model

Sora 2 Pro. Tell a whole story in one clip.

OpenAI's flagship video model generates the longest clips in the Treza lineup, with synced dialogue and a strong sense of scene continuity. One prompt, a complete beat.

  • 4 to 20 second clips
  • 16:9 and 9:16
  • Native audio
The model

Where most models give you a shot, Sora 2 Pro gives you a scene. Clips run up to 20 seconds, long enough for a setup and a payoff, and the model keeps characters, props, and lighting coherent as the action unfolds. Dialogue and effects are generated in sync, so a scripted line lands on camera.

That length changes what you can build on Treza. A single Sora 2 Pro node can carry a full product story or a spoken testimonial, and a pipeline can chain a language model that writes the beat sheet straight into the generation. Publish the pipeline and your product gets a story-length video API.

Up to 20 second clips

The longest single generations in the lineup. Enough room for a setup, a turn, and a close without stitching.

Scene continuity

Characters, wardrobe, and lighting stay coherent across the clip as the camera and action move.

Synced dialogue

Write the line in the prompt and it is spoken on camera, with ambience and effects generated around it.

Strong instruction following

Multi-part prompts with staging, timing, and camera notes are followed closely, which makes long clips directable.

Prompt to result

What to make with Sora 2 Pro

Real prompts, and what comes back. Every one runs in Treza chat or as a node on the canvas.

A 20 second ad: a runner struggles up a hill, reaches the top at sunrise, tagline spoken over the final wide shot

A complete narrative ad beat in a single generation.

A chef plates a dish while explaining it to camera, kitchen ambience, one continuous take

A talking segment with synced speech and kitchen sound.

A robot wakes up in a workshop, looks around, picks up a tool, 16 seconds

A character beat with coherent staging start to finish.

A phone slides across a desk, rings, gets picked up, caller speaks

Sequenced actions with sound landing on each beat.

A vertical 12 second explainer: rain starts, umbrella opens over the product

A portrait story clip ready for feeds.

Two colleagues exchange a short scripted dialogue in an elevator

A two-person scene with alternating spoken lines.

Beyond one prompt

Sora 2 Pro is one node on a canvas

Chain it with language, image, and video models, then publish the whole pipeline as an API.

  1. Step 01

    Draft the brief

    A language model node expands a one-line idea into a detailed, on-brand prompt before it ever reaches the model.

  2. Step 02

    Generate with Sora 2 Pro

    Run variants in parallel, compare against other models on the same prompt, and keep every result in run history.

  3. Step 03

    Publish it as an API

    The pipeline becomes a versioned endpoint. Your product calls it with the typed /invoke API or any OpenAI SDK.

FAQ

Sora 2 Pro, answered

How long can Sora 2 Pro videos be?

Sora 2 Pro generates clips up to 20 seconds long on Treza, the longest single generation in the lineup. That is enough for a complete narrative beat without editing clips together, and you can go longer by chaining generations in the Treza editor.

Can Sora 2 Pro handle dialogue?

Yes. Scripted lines in the prompt are spoken on camera with synced audio, and multi-character exchanges work in a single clip.

When would I pick Sora 2 Pro over Veo 3.1?

Pick Sora 2 Pro when the idea needs duration and storytelling: sequences, spoken segments, setups with payoffs. Pick Veo 3.1 when a single shot needs maximum cinematic polish. Both sit in the same Treza model picker, so trying each on the same prompt costs one credit run.

How much does Sora 2 Pro cost on Treza?

Treza uses prepaid credits with no subscription. Each clip is charged at Sora 2 Pro's rate from your balance, and only successful generations are charged. Credit packs start at $5 and never expire.

Can I call Sora 2 Pro through an API?

Yes. Build a pipeline that uses Sora 2 Pro and publish it as a versioned HTTP endpoint. Call the typed /invoke API from your product, or point any OpenAI SDK at the OpenAI-compatible endpoint. If a better model ships later, swap it on the node without changing the API your product calls.