Workflow Optimization ·

Sora 2 Prompt “Containers” vs. Prompts: The 15-Minute Creator Workflow to Stop Ignored Duration, Resolution, and Character Settings (as of 2026-04-13)

Stop “duration/resolution/character” being ignored in Sora 2: use container parameters first, then write the shot. A 15‑minute workflow + examples.

On this page

Why your prompt gets ignored: “container parameters” vs. prompt text

If you’ve ever written “make it 20 seconds, 4K, same character” and the output still comes back short, low-res, or with a slightly different face—this is usually not a “bad prompt” problem. It’s a where you put the instruction problem.

In plain English:

  • Container parameters are the generation settings that surround your prompt—think: the form fields / API parameters that define the boundaries of the clip.
  • Prompt text is the shot description—what happens inside those boundaries.

OpenAI’s Sora 2 Prompting Guide is explicit that certain video attributes are governed only by API parameters and cannot be reliably requested in prose. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

As of 2026-04-13, the guide calls out capabilities that map neatly to “containers,” including:

When you treat those like “prompt adjectives” instead of hard settings, your prose can be ignored—because it’s asking the model to break its own container.

The 15-minute creator workflow: set containers first, then write the shot

This workflow is designed for solo creators who want repeatable outputs—marketing clips, social ads, explainers, storyboards—without endlessly re-prompting.

Step 1 (3 minutes): Pick your model container

Decide which model family you’re targeting (for Sora API, the guide lists sora-2 and sora-2-pro). (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

This matters because supported sizes differ by model. For example, the guide lists sora-2 supporting 720x1280 and 1280x720. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

And it lists sora-2-pro supporting additional sizes including 1920x1080 and 1080x1920. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

Step 2 (3 minutes): Lock duration and aspect

Choose your duration as a container value, not a sentence:

The guide notes that maximum duration increased from 12 seconds to 20 seconds. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

Then choose size using the {width}x{height} string format. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

Step 3 (4 minutes): Decide character consistency up front

If your clip needs the “same person” across variants or across episodes, treat character identity as a container reference:

Step 4 (5 minutes): Write the shot prompt (only what belongs inside the container)

Now write your prompt text as if duration/resolution/identity are already handled.

Focus on:

  • Subject + action
  • Location + time of day
  • Camera/shot language (wide/close, movement)
  • Lighting + style references
  • What must not happen (sparingly)

One short checklist (under 15 minutes)

Copy/paste “Container Card” (decide this for every clip)

Use this as a preflight card in your notes app or production doc.

Container Card (Sora 2-style)

Copy/paste “Shot Card” (what to write in the prompt every time)

Shot Card (Prompt Text)

  1. One-line intent: What should the viewer feel/understand?
  2. Subject & wardrobe: Who/what, key visual traits (but not “same character” if you’re using references).
  3. Setting: Location, era, time of day, weather.
  4. Action beats: 2–4 sequential beats (what happens).
  5. Camera language: lens vibe, framing, movement.
  6. Lighting & color: soft/hard, warm/cool, contrast.
  7. Style constraints: e.g., “documentary realism,” “product commercial,” “handheld street footage.”
  8. Guardrails: 1–2 “avoid” statements (avoid logo distortion, avoid extra limbs, etc.).

Common failure modes (and the exact fix)

Duration ignored (“make it longer”)

Symptom: You ask for “longer,” but you still get a short clip.

Fix: Set seconds to the target value (16 or 20) instead of requesting duration in prose. The guide lists supported seconds values and notes certain attributes must be set via parameters. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

If you need a story that exceeds a single clip, plan to extend a clip; the guide describes extension using the full initial clip as context. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

Resolution ignored (“4K”, “1080p”, “vertical”)

Symptom: You request “4K” or “vertical,” but exports don’t match.

Fix: Set size explicitly using {width}x{height} supported by your model. For sora-2-pro, the guide lists sizes including 1920x1080 and 1080x1920. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

Character consistency drifts (“same character as before”)

Symptom: Same “person” changes hair/face/clothes across takes.

Fix: Use character references. The guide says you can upload a character once and reuse it across videos with consistent appearance. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

Mini decision tree: change containers vs rewrite the prompt

Example: one marketing clip rewritten (bad prompt vs fixed setup + prompt)

Bad prompt (why it fails)

“Make a 20-second 4K vertical ad for a new smart water bottle. Same female athlete character as before. Cinematic lighting. Make it longer if needed.”

What’s wrong: It mixes container requirements (duration, resolution, vertical format, character reuse) into prose. The Sora guide warns some attributes are controlled only by parameters, not by text. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

Fixed setup + prompt (containers first)

Container Card

Shot prompt (text only)

“Vertical product commercial. A focused female athlete finishes a sunrise run on a quiet city bridge. She stops, pulls a sleek smart water bottle from her bag, and taps the cap; a subtle indicator light turns on. Close-up of condensation and water swirling as she drinks. Cut to a medium shot as she smiles, stretches, and jogs away. Cinematic golden-hour lighting, shallow depth of field, smooth gimbal movement, clean modern color grade. Keep branding minimal and legible; avoid warped text.”

This split keeps the model from negotiating with your “please be longer” sentence—because the clip’s boundaries are already decided.

How to port this thinking to Veo3Gen (so settings don’t fight your prompt)

Even if your tool UI or API doesn’t match Sora 2 parameter names 1:1, the mental model transfers:

  1. Set generation settings first (duration, resolution/aspect, reference images/characters where available).
  2. Then write the shot (action, camera, style, guardrails).

The win is practical: you stop overloading the prompt with “meta” instructions that belong in settings, and you get cleaner iteration loops—especially when you’re producing multiple variations for the same campaign.

Quick FAQ (as of 2026-04-13)

What are “container parameters” in Sora 2?

They’re the settings around the prompt that define constraints like duration (seconds), resolution (size), and character reuse (characters)—attributes that the guide notes aren’t reliably controlled via prose. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

Can I ask for 1080×1920 just by writing “vertical 1080p”?

The guide describes setting size as {width}x{height} and lists supported resolutions by model; for sora-2-pro, 1080x1920 is supported. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

What durations can I set?

The guide lists seconds values 4, 8, 12, 16, 20 (default 4) and notes the maximum increased to 20 seconds. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

How do I keep the same character across clips?

Use character references: upload the character once, then reference its ID in the characters parameter (up to two per generation). (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

If I need something longer than one clip, what should I do?

Plan a multi-clip workflow and use video extension where appropriate; the guide states extension can use the full initial clip as context. (https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)

CTA: turn this workflow into repeatable generation

If you want to operationalize the “containers first, shot second” approach in your own pipeline, Veo3Gen is built for structured generation workflows.

  • Explore the endpoints and how to pass generation settings cleanly: /api
  • Estimate costs and plan production runs: /pricing

Try Veo3Gen (Affordable Veo 3.1 Access)

If you want to turn these tips into real clips today, try Veo3Gen:

  • Start generating via the API: /api
  • See plans and pricing: /pricing

Sources

Limited Time Offer

Try Veo 3 & Veo 3 API for Free

Experience cinematic AI video generation at the industry's lowest price point. No credit card required to start.

Get Started FreeView Cheapest Pricing
← Back to all posts