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.
Q3 Turbo
Q3 Turbo
Q3 Turbo
External tests and provider examples
These are reports from the named publishers and providers; FrameSignal did not run these generations.
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 ↗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 ↗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 ↗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 ↗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 ↗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 ↗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