Video generation cluster: Runway Gen-4.5 + Veo 3.5 + Sora 2 + Pika 2.2 — every video model ships a longer-coherent / native-audio milestone in one week

6 sources

The week of 2026-08-08 saw a cluster of video generation model releases: Runway Gen-4.5 (motion coherence over 5+ seconds + character / object consistency), Google DeepMind Veo 3.5 (4K + native audio + camera-motion API on Vertex AI), OpenAI Sora 2 (60+ second shots + native audio + Storyboard editor), Pika 2.2 (Pikaframes keyframe-to-keyframe interpolation + Pika Additions in-painting API). The Decoder frames the cluster as 'every video model ships a longer-coherent / native-audio milestone in the same week'. The pattern: video generation is converging on two axes — longer temporal coherence and native audio generation.

Why now

The cluster is the editorial framing that turns four independent video model releases into a single coherent 'video generation converges on coherence + audio' story for creators.

Why it is worth publishing

Demo potential: side-by-side of the same 30-second storyboard generated across all four models, measuring motion coherence and audio fidelity.

Evidence basis

The Decoder + IT之家 weekly roundups + four independent vendor primary sources

“Four video generation models shipped in one week — Runway Gen-4.5, Veo 3.5, Sora 2, Pika 2.2 — and every one shipped a longer-coherent shot or native-audio milestone.”

Angle

Use the cluster to introduce the 'video generation converges on coherence + audio' pattern — every vendor shipped longer coherent shots + native audio in the same week — and use that lens to compare models side-by-side.

Format

Long-form explainer

Demo idea

Record a 14-minute comparison explainer: 2 min intro on 'coherence + audio' framing, then 2 min per model (Runway / Veo / Sora / Pika), then a 4-min side-by-side of the same 30-second storyboard across all four.

Platform notes

Each vendor's release notes frame its release against the competitive set it cares about; The Decoder and IT之家 are editorial framing layers, not independent verification. Motion coherence and character consistency claims are vendor self-reported; confirm with creator-side demo.

Usable claims

Runway Gen-4.5 video generation model reached GA — improved motion coherence over 5+ second shots, character / object consistency across scenes, Act-Two motion-capture upgrade.
Google DeepMind Veo 3.5 GA on Vertex AI — 4K resolution support, native audio generation (lip-synced dialogue, ambient sound), camera-motion API.
OpenAI Sora 2 — 60+ second shots with character / scene consistency, native audio generation, Sora Storyboard editor.
Pika 2.2 — Pikaframes keyframe-to-keyframe interpolation (specify start and end frames), improved scene consistency, Pika Additions API for in-painting.

Evidence pipeline

From the news

Breakdown

Four video models shipped in one week — the editorial framing ('every video model ships a longer-coherent / native-audio milestone') is useful but turns into a spec sheet if you don't introduce the convergence pattern. This explainer uses the cluster to introduce the 'video generation converges on coherence + audio' pattern and uses that lens to compare models side-by-side, while keeping the per-model differences honest (Runway on motion coherence + Act-Two, Veo 3.5 on 4K + Vertex AI integration + camera-motion API, Sora 2 on 60+ second long-take + Storyboard editor, Pika 2.2 on Pikaframes keyframe interpolation + Additions API).

Sources

Risks

Use The Decoder and IT之家 as media-type corroboration, but read the underlying vendor docs for any specific capability or pricing claim before stating it on the record. Verify specific capability claim against the underlying vendor docs and the actual license / pricing matrix before stating it on the record; do not paraphrase per-platform pricing or license terms into specific dollar figures or commercial-use clauses.
Vendor docs confirm feature existence but specific pricing beyond the captured summary is not in this pass. Verify specific capability claim against the underlying vendor docs and the actual license / pricing matrix before stating it on the record; do not paraphrase per-platform pricing or license terms into specific dollar figures or commercial-use clauses.
Vendor docs describe motion coherence as a release-note property; creator-side demo is the right way to verify. Verify specific capability claim against the underlying vendor docs and the actual license / pricing matrix before stating it on the record; do not paraphrase per-platform pricing or license terms into specific dollar figures or commercial-use clauses.

Demo ideas

Side-by-side 30-second storyboard: same prompt on all four models, measure motion coherence and audio fidelity.
Decision tree: 'which model for which use case' (motion-capture-driven → Runway Act-Two, 4K + native audio on Vertex → Veo 3.5, 60+ second shots → Sora 2, keyframe-directorial → Pika Pikaframes).
Coherence vs audio matrix: plot each model on temporal coherence vs native-audio fidelity.