Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
TikTok's 4-tier penalty system is now enforced via C2PA Content Credentials, which auto-flag synthetic media even when creators don't disclose. Unlabeled AI content catches a minus-73% reach penalty within 48 hours of posting. The penalty ladder runs warning, 7-day restriction, 30-day suspension, permanent ban — and it applies to synthetic faces, voice clones, AI backgrounds, and photorealistic AI products.
Hailuo 2.3 is benchmarking top of WorldModelBench for physics simulation and generating 30-to-90-second clips at roughly $0.30 per 10-second 1080p clip — vs $1.20 on Runway and $3.50 on Sora. Fast variants cut batch costs another 50% for volume operators. At that price delta, the cost math on a 50-clip production week shifts significantly.
The most actively updated all-in-one ComfyUI video workflow on Civitai just hit v3.0, integrating LTX 2.3 distilled (better emotion and audio sync), Wan 2.2 I2V, Stable Character Voice, and T2V/I2V/A2V in a single flow. LTX 2.3 outputs 4K with native stereo audio at 24kHz — added in March 2026. Low-VRAM variants are working on RTX 3060 12GB and down to 8GB cards.
Alibaba's Wan 2.5 and 2.6 — which added multi-shot prompts and reference-to-video — are closed API-only releases. Wan 2.2 remains the open local option and is still the ComfyUI workhorse for operators running on-prem. If you're building local pipelines, Wan 2.2 is likely your last native checkpoint for this family — Alibaba appears to be moving the roadmap behind the API wall.
Under Oracle/US ownership, TikTok's algorithm now weights watch time over raw view counts, rewards niche-specific content, and uses first-hour engagement signals as the primary distribution trigger. Optimal length has shifted to 60-180 seconds — the sub-30-second clip era is functionally over for maximizing FYP push. Users can now manually filter AI content from their feed, which means low-quality volume plays have a shrinking ceiling.
YouTube Shorts ad revenue runs $0.01-$0.12 per thousand views at 45% creator share — finance and tech niches hit $0.10-$0.20+ RPM while entertainment and comedy land at $0.01-$0.03. Long-form RPM is 10x to 50x higher in the same niche. AI tutorial channels are pulling $8-$15 RPM on long-form; generic AI faceless channels are stuck at $1.50-$3.00 RPM. The play is Shorts as top-of-funnel, long-form as the monetization engine.
The dominant production pattern across the creator community right now is one concept rendered into vertical, horizontal, short-form, and localized language variants in a single pipeline pass. The capability gaps being actively solved: character consistency across scenes, synced audio on generated clips, in-clip scene edits without full regen, and multi-language output. Operators building that stack are pulling ahead on content volume without proportionally higher cost.
The "which would you choose" prompt format is racking up outsized engagement right now — one recent example clocked 20.7M views, 1M likes, 17.1K comments, and 41.1K saves on a single post. The mechanic is simple: present two AI-generated scenarios and ask viewers to pick, which drives comment spikes and saves. It's a format that pairs naturally with AI video because the visual contrast is the whole hook.
