Evidence review · 2026

Vidu Q3 Turbo review: verified public evidence

This evidence review is based on official records, provider data, independent public tests, and external benchmarks. It does not claim FrameSignal hands-on testing.

Updated Sep 23, 2026Model brief
External evidence summary. A FrameSignal hands-on review has not been completed. The source reports below are also included in the model details.

Current evidence summary

Vidu Q3 Turbo supports text-to-video, image-to-video, start/end-frame and reference-to-video workflows with synchronized generated audio, including all-audio, speech-only and sound-effects-only modes. Compared with Q2 Turbo, Q3 Turbo adds text-to-video and reference-to-video and extends main-route clips up to 16 seconds. Vidu's official live price table currently lists 12 credits per 720p generated second, while its documentation pricing page still lists 11 credits ($0.055); both values are shown with their source and the conflict remains unresolved. Provider prices and matched-prompt completion times vary materially by route. Vidu Q3 Turbo is distinct from the separately tracked Q3 Drama model. The source-backed use cases currently point to native audio text-to-video and image-to-video workflows, first and last frame control or reference-to-video generation, longer q3 turbo clips up to 16 seconds, teams comparing vidu credits with third-party per-second routes and actual job latency.

Public benchmark snapshots

Each result stays attached to its task, date, sample count, and reported uncertainty. Developer backend timings are kept separate from end-to-end wait tests.

Text to Video Leaderboard · Sep 23, 2026

Q3 Turbo

text to videoElo1025 Elo · rank 28 · n=3,117
text to videorank28 rank · rank 28 · n=3,117
text to videosamples3117 comparisons · rank 28 · n=3,117
Open benchmark source ↗
Image to Video Leaderboard · Sep 23, 2026

Q3 Turbo

image to videoElo1030 Elo · rank 25 · n=3,267
image to videorank25 rank · rank 25 · n=3,267
image to videosamples3267 comparisons · rank 25 · n=3,267
Open benchmark source ↗
Model Detail Quality Analysis · Sep 23, 2026

Q3 Turbo

aggregate or model detailquality elo1105.98 Elo
Open benchmark source ↗

External tests and provider examples

These are reports from the named publishers and providers; FrameSignal did not run these generations.

Artificial Analysis · Artificial Analysis · reliability A

matched prompt provider latency

Reported metric: end to end time — fal.ai: 45.6s; Replicate: 84.1s seconds

Same prompt/provider comparison; timing varies by run.

Reported result: Same prompt/provider comparison; timing varies by run.

Prompt: Sculptor hovering in a botanical garden; handheld sway; desaturated travelogue.

Settings: target: 720p 5s benchmark configuration · providers: ["fal.ai","Replicate"]

Open source ↗
Artificial Analysis · Artificial Analysis · reliability A

matched prompt provider latency

Reported metric: end to end time — Replicate: 55.6s; fal.ai: 56.7s seconds

Same prompt/provider comparison; timing varies by run.

Reported result: Same prompt/provider comparison; timing varies by run.

Prompt: Border collie in autumn orchard; dappled light; orbiting crane shot.

Settings: target: 720p 5s benchmark configuration · providers: ["Replicate","fal.ai"]

Open source ↗
Artificial Analysis · Artificial Analysis · reliability A

matched prompt provider latency

Reported metric: end to end time — fal.ai: 56.8s; Replicate: 187.9s seconds

Same prompt/provider comparison; timing varies by run.

Reported result: Same prompt/provider comparison; timing varies by run.

Prompt: Potter at a flickering night market; slow pan; anamorphic film.

Settings: target: 720p 5s benchmark configuration · providers: ["fal.ai","Replicate"]

Open source ↗
Artificial Analysis · Artificial Analysis · reliability A

matched prompt provider latency

Reported metric: end to end time — fal.ai: 47s; Replicate: 50.2s seconds

Same prompt/provider comparison; timing varies by run.

Reported result: Same prompt/provider comparison; timing varies by run.

Prompt: Calligrapher on a sand dune; slow tilt; clay stop-motion.

Settings: target: 720p 5s benchmark configuration · providers: ["fal.ai","Replicate"]

Open source ↗
Artificial Analysis · Artificial Analysis · reliability A

matched prompt provider latency

Reported metric: end to end time — fal.ai: 49.1s; Replicate: 68.9s seconds

Same prompt/provider comparison; timing varies by run.

Reported result: Same prompt/provider comparison; timing varies by run.

Prompt: Cyclist spiraling above a quiet river; overhead top-down; clay stop-motion.

Settings: target: 720p 5s benchmark configuration · providers: ["fal.ai","Replicate"]

Open source ↗
Runware · Runware · reliability A

text to video product demo

Reported metric: raw generation charge — 0.2145 USD

Provider shows roughly 48–65s end-to-end for 6s 720p T2V examples.

Reported result: Provider shows roughly 48–65s end-to-end for 6s 720p T2V examples.

Prompt: Aquaculture net-cleaning robot product demonstration; full prompt at source.

Settings: duration seconds: 6 · resolution: 720p · audio: true

Open source ↗
Runware · Runware · reliability A

first last frame

Reported metric: raw generation charge — 0.17875 USD

Provider shows about 3m21s end-to-end for this example.

Reported result: Provider shows about 3m21s end-to-end for this example.

Prompt: Logo reveal between supplied opening and closing frames; full prompt at source.

Settings: duration seconds: 5 · mode: first_last_frame

Open source ↗

What official evidence suggests

Promising areas

  • Official docs cover T2V, I2V, start/end-frame and reference-to-video
  • Native synchronized audio supports all, speech-only and sound-effects-only modes
  • Canonical 540p, 720p and 1080p output with up to 16 seconds on main routes
  • Current matched-prompt provider runs expose substantial route-level latency variation

Verify before paying

  • Vidu official pricing surfaces conflict on general 720p at 11 versus 12 credits per second
  • Matched-prompt timings are provider end-to-end times and vary by run
  • Runware's 360p listing is provider-specific and not a canonical Vidu resolution
  • The 1105.98 model-detail quality Elo is a different scope from T2V and I2V leaderboard Elo
  • Generalized V2V and editing remain unverified for this variant