Research profile · Sora 2 (December) and Sora 2 Pro
Sora 2, researched.
Highest blind-preference placement in this set. Overall verdict withheld. Explore the findings, limitations and 3 cited sources below.
We are withholding an overall verdict on Sora 2. Evidence coverage is 17%, below the 75% our method requires. Highest blind-preference placement here, but with no standalone price and no published consumer limits, most criteria stay open.
- Researched scope
- Sora 2 (December) and Sora 2 Pro
- Best supported for
- Prompt-faithful scenes with generated audio, if you already pay for ChatGPT
- Principal limitation
- No standalone pricing and no published consumer generation limits
- Accepted sources
- 3 · 2 of 12 criteria supported
Best for, and not for
Best supported use
Prompt-faithful scenes with generated audio, if you already pay for ChatGPT. The strongest finding is highest blind-preference placement in this set.
Where the evidence warns
No standalone pricing and no published consumer generation limits.
What the evidence supports, criterion by criterion
2 of 12 applicable criteria have enough accepted evidence to say something. The rest are listed below them, unscored.
| Criterion | Evidence | What we found |
|---|---|---|
| Visual fidelity and aesthetics | Limited evidence | Sora 2 (December) is the highest-placed model in this set on blind preference, Elo 1097 from 4,832 votes, with Sora 2 Pro at 1076.Vote count is under a third of Vidu's, so the interval is correspondingly wider. |
| Native audio and lip sync | Limited evidence | Native synchronised audio is a headline capability of the Sora 2 announcement, and one of the few models in this set where audio is generated with the video rather than added after.Vendor announcement only; no independent evaluation of the audio was accepted. |
Not established
Researched, but the available public evidence does not support a conclusion. These are gaps, not low scores.
- Temporal stability and artifact control — Researched; no source meeting our weighting bar addressed this for the scoped version.
- Prompt or script adherence and control — Researched; no source meeting our weighting bar addressed this for the scoped version.
- Reliability and iteration burden — Researched; no source meeting our weighting bar addressed this for the scoped version.
- Workflow and editability — Researched; no source meeting our weighting bar addressed this for the scoped version.
- Output specifications and delivery — Researched; no source meeting our weighting bar addressed this for the scoped version.
- Pricing and value — Cost per clip not derivable. There is no standalone price: access is bundled into paid ChatGPT tiers, and OpenAI has not published per-generation limits for consumer tiers.
- Camera motion — Researched; no source meeting our weighting bar addressed this for the scoped version.
- Character and object consistency — Researched; no source meeting our weighting bar addressed this for the scoped version.
- Physics and material realism — Researched; no source meeting our weighting bar addressed this for the scoped version.
- Text rendering — Researched; no source meeting our weighting bar addressed this for the scoped version.
Pricing
Published list price, normalised to a five-second generation. Assumes full use of the plan allocation, before tax, no retakes — so real cost per clip you keep is higher.
| Plan | Cost / 5s | Resolution | Audio |
|---|
All Sora 2 pricing scenarios and how they were normalised. Prices verified 2026-09-12 and change without notice.
Alternatives
Same category, so genuinely comparable. We do not rank across categories.
- Kling AI — Complex physical movement and holding a character's face steady across a shot. Moderately supported.
- Higgsfield — Camera-move control and stylised cinematic shots, with access to several models in one place. Not established.
- Runway Gen-4.5 — Creative and editorial work — generative editing, motion brush and in-context controls. Limited evidence.
Sources
Every finding above traces to one of these. Version match records how closely each source matches Sora 2 (December) and Sora 2 Pro.
| Source | Type | Version match | Accessed |
|---|---|---|---|
| OpenAI | Official documentation | Exact | 2026-09-12 |
| OpenAI | Official documentation | Exact | 2026-09-12 |
| Artificial Analysis | Independent benchmark | Exact | 2026-09-12 |