Vidu Q4 Preview Debuts at No. 3 on AI Video Leaderboard

👁 115 reads

Vidu has vaulted back into the top tier of AI video generators. Released on October 7, 2026, the Vidu Q4 Preview model debuted in 3rd place on the Artificial Analysis image-to-video leaderboard with an Elo rating of 1,179 points. The company’s previous flagship, Vidu Q3 Pro, sat at #19. Only two MiniMax models rank higher, with just a two-point gap separating Vidu from second place. I have gathered all numbers, pricing details, and tutorials in the Vidu Q4 guide.

What Vidu Delivered

ShengShu Technology, a company founded by Tsinghua University researchers, announced Vidu Q4 Preview on October 7 in Singapore. This is the first public iteration of their new flagship model, rolled out ahead of the complete Vidu Q4 release. The team wants creators to deploy it in real-world production before finalizing the full build. Key specifications:

• Clips from 3 to 16 seconds, in resolutions from 540p up to 4K, featuring 10-bit color in 2K and 4K;
• Native audio: dialogue, sound effects, and ambient background generated concurrently with video frames;
• Two generation modes: image-to-video (single starting frame) and reference-to-video (up to 15 reference images and up to 3 voice samples for tone cloning);
• No text-to-video mode: generations must begin with an input image.

Vidu highlights improvements across three core areas: character acting (synchronized facial expressions, emotion, and vocal delivery), camera movement during fast action sequences, and visual effects designed to integrate seamlessly into scenes. ShengShu CEO Yihang Luo noted in the release that cutting-edge video models must prove themselves beyond curated demonstration reels.

Ranked #3, Effectively a Dead Heat at the Top

The Artificial Analysis leaderboard relies on blind side-by-side evaluations: human raters inspect two clips generated from identical inputs and vote for the superior render. Vidu Q4 Preview logged 5,543 votes within its first 24 hours. The top image-to-video standings as of October 8:

A mere 6 points separates 2nd place from 6th place, while Vidu Q4 carries a confidence interval of plus or minus 10 points. Statistically speaking, the five models directly trailing H3 Max sit in a tie. For Vidu, this leap represents a gain of 16 ranks and 123 Elo points over Q3 Pro. Artificial Analysis also published side-by-side output comparisons on identical inputs in an X thread. The full top 13 leaderboard is detailed in the Vidu Q4 guide.

Pricing and Availability

API pricing matches Vidu Q3 Pro rates: 24 credits per second of 1080p footage with audio, equaling $0.12 per second or $7.20 per minute at standard rates. 720p runs at $0.095 per second, while 4K costs $0.39 per second. The advertised baseline of “from $0.014 per second” applies only to the lowest resolution under initial promotional discounts. Current access points as of October 8:

• vidu.com - paid subscription tiers starting at $10/month (800 credits) alongside a free tier featuring a sign-up credit bonus; Q4 at 540p costs 2 credits per second on paid plans and 4 credits on the free tier;
• Vidu API - endpoint model ID viduq4-preview;
• fal.ai - 30% discount through November 30 ($0.084 per second for 1080p);
• WaveSpeedAI and partnered ComfyUI nodes.

Vidu Q4 has not yet arrived on OpenRouter or Replicate. Over on reAPI, access remains limited to Vidu Q3.

Industry Impact

For weeks, the image-to-video leaderboard appeared to be a closed race between MiniMax, Google, and ByteDance. Vidu Q4 introduces 4K capability and voice cloning into the top echelon, all while keeping 1080p API costs below Seedance, Wan 3.0, Kling, and Veo. Its main drawback remains the lack of pure text-to-video, meaning projects relying entirely on prompt descriptions are still better served by Wan 3.0 or MiniMax H3. I have not generated custom test clips from Q4 yet; a five-scene evaluation suite is ready, and those results will be added directly to the Video Arena. For head-to-head performance breakdowns, consult Vidu Q4 vs Seedance 2.0 vs MiniMax H3, or follow the step-by-step setup in how to generate video from images in Vidu Q4.

Sources: Vidu press release (Oct 7, 2026), Artificial Analysis leaderboard (Oct 8, 2026), Vidu API pricing, vidu.com pricing plans, fal.ai.

// read moreVidu Q4 Preview: AI Video Benchmark Rank, Pricing, and GuideAI ToolsVidu Q4 Image to Video: Prompts, API, and Full WorkflowGuidesSeedance 2.5: Video Leaderboard Podium and Real API CostsBenchmarks