Research profile · Q3 Pro, and the faster Q3 tier (Turbo in API docs, Flash on the pricing page)
Vidu, researched.
Transparent published pricing and the largest arena sample in this set. Provisional findings. Explore the findings, limitations and 7 cited sources below.
Vidu Q3 is supported for reference-driven work with the most transparent pricing in the set and the largest arena sample. On the one academic benchmark that measures it, the Turbo tier is mid-table on motion and cinematography, weak on physics, and shows the same collapse from entity to action to cause seen across the field. Every quality finding is on Turbo; Pro is unmeasured, and the vendor's own pages cannot agree on whether Turbo and Flash are the same tier.
- Researched scope
- Q3 Pro, and the faster Q3 tier (Turbo in API docs, Flash on the pricing page)
- Best supported for
- Reference-to-video where a supplied subject must persist, on a published per-clip price
- Principal limitation
- Action, state and causal prompts; physics plausibility; the Pro tier is unmeasured
- Accepted sources
- 7 · 8 of 11 criteria supported
Best for, and not for
Best supported use
Reference-to-video where a supplied subject must persist, on a published per-clip price. The strongest finding is transparent published pricing and the largest arena sample in this set.
Where the evidence warns
Action, state and causal prompts; physics plausibility; the Pro tier is unmeasured.
What the evidence supports, criterion by criterion
8 of 11 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 | Q3 Pro scores Elo 1075 from 18,849 blind votes — the largest vote count of any tool in this set, and closely matched with PixVerse V6 and Sora 2 Pro.Preference Elo does not decompose into specific quality attributes. |
| Temporal stability and artifact control | Limited evidence | Q3 Turbo scores 46.19 on VGIF-Score's motion-smoothness rubric, on a 0-100 scale where fourteen models span 26.28 to 55.76 — mid-table, and a little above PixVerse V6.Automatic rubric judgment rather than human raters, and it measures the Turbo tier; whether Turbo is the same thing as the pricing page's 'Flash' is still unmapped. |
| Prompt or script adherence and control | Conflicting evidence | The same pattern seen on PixVerse: Q3 Turbo holds what is in the shot and loses what happens in it. Entity 74.35 and location 78.73, against action 35.57, state 10.95 and causal 3.68, across 223 prompts.The pattern is the finding; the absolute figures are not, for the same reasons as PixVerse — the paper's composite does not reconcile with its repository normalisation. Turbo tier; Flash mapping unresolved. |
| Workflow and editability | Moderately supported | Q3 Pro documents text, image, reference and start-end-frame routes, with audiovisual sync and storyboard generation on Pro. The documented video-extension endpoint accepts only Q2 models, so extending Q3 output is unsupported there.Feature documentation only. Vidu names its faster Q3 tier Turbo in the API docs and Flash on the pricing page; we have not found a vendor mapping between the two and do not assume they are the same. |
| Output specifications and delivery | Moderately supported | Official docs specify up to 16 seconds at 1080p across text-to-video, image-to-video, reference-to-video and start-end-frame modes.Documentation establishes the offering, not performance. |
| Pricing and value | Moderately supported | Fully published credit table: Q3 Flash 1080p text-to-video runs 10 credits for 1-5s up to 60 for 15-16s; Cinematic 1080p runs 40 to 140 over the same range. 40 free credits monthly.Among the most transparent pricing in the set; per-clip value still depends on unmeasured retry rates. |
| Camera motion | Limited evidence | Cinematography scores 44.93 for Q3 Turbo on VGIF-Score's rubric, which bundles camera work with composition, lighting and pacing.A composite, not isolated camera-motion accuracy, and the weakest fit of the two new criteria. Turbo tier; Flash mapping still unresolved after five rounds. |
| Physics and material realism | Limited evidence | VGIF-Score's Physics Adherence rubric — plausibility of requested interactions and state changes — scores Q3 Turbo at 34.35, on a scale where the same table's leader sits in the mid-50s.A language-model rubric judgment, not a physical measurement. Turbo tier; whether Pro or Flash differ is unmeasured. The table's scale is presentational. |
Not established
Researched, but the available public evidence does not support a conclusion. These are gaps, not low scores.
- Character and object consistency — VGIF-Score's Visual Purity dimension (67.62) bundles identity drift with artefacts and extraneous elements and was correctly refused as a consistency measure. A Zhihu comparison scoring 'Vidu 4.9/10' on facial consistency names no release or tier and could not be assigned to Q3.
- Text rendering — Searched Shengshu's Q3 documentation and technical material on 14 September 2026; no statement on text rendering was found. VGIF-Score does not measure it.
- Native audio and lip sync — Searched on 14 September 2026 including Chinese-language sources; Sohu and Baidu Baike describe 16-second audiovisual generation as a capability but nothing measures its sync. VGIF-Score does not cover audio.
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 Vidu pricing scenarios and how they were normalised. Prices verified 2026-09-14 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. Moderately supported.
Sources
Every finding above traces to one of these. Version match records how closely each source matches Q3 Pro, and the faster Q3 tier (Turbo in API docs, Flash on the pricing page).
| Source | Type | Version match | Accessed |
|---|---|---|---|
| Shengshu | Official documentation | Exact | 2026-09-12 |
| Shengshu | Official documentation | Exact | 2026-09-12 |
| Artificial Analysis | Independent benchmark | Exact | 2026-09-12 |
| Shengshu | Official documentation | Exact | 2026-09-13 |
| Shengshu | Official documentation | Exact | 2026-09-13 |
| Shengshu | Official documentation | Exact | 2026-09-13 |
| Songyu Xu et al. (PRCV 2026) | Academic evaluation | Exact | 2026-09-14 |