Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
TikTok's algorithm is suppressing static AI avatars by 47%, tanking share rates on template AI voiceovers by 58%, and driving 62% higher abandonment on low-framerate AI product demos — those numbers are from TikTok's own 2026 creator report. The fix is motion and human signal: dynamic camera movement, genuine environmental variation, or human footage layered with AI enhancements. Pure synthetic pipelines are getting clipped hard.
Betrayal and revenge narratives are hitting $12.82 RPM ($20-25 CPM), English learning podcasts at $11.88 RPM, soundscapes for sleep and healing at $10.92 RPM, and manhwa/webtoon recaps at $10.45 RPM — all verified from creator dashboard data and all flagged as AI-narrated faceless formats. Manhwa recaps are the sleeper: only 10K competing channels with repeatable source material and a $15-19 CPM ad market.
YouTube's own 2026 internal data shows creators posting both Shorts and long-form grow subscribers 3x faster than single-format operators. The RPM gap is brutal — Shorts earns $0.01-$0.07 per 1K views vs $3-$20+ on long-form in the same niche, a 15-60x difference. The play is using AI Shorts as top-of-funnel traffic drivers with pinned comments and end screens pointing to longer monetized content.
TikTok Shop hit $23.4 billion US GMV and is nearly doubling year-over-year. The winning operator framework is 70% human-created published content (keeps monetization eligibility) plus 30% AI-powered for hook testing, brief generation, and pattern mining — that 30% never needs labeling because it's text-only. On the production side, HeyGen drops AI UGC to roughly $1.50 per video vs $80 for human marketplace talent. Superscale's January 2026 research clocks AI content at 350% higher engagement on TikTok but 18 points lower on authenticity trust — hence why you keep humans on camera.
Alibaba dropped Wan 2.7 into ComfyUI on April 3rd via Partner Nodes — five task types including reference-to-video with up to 5 real-person image inputs plus vocal timbre referencing, audio-driven image-to-video, video continuation, and text-driven video editing. Update to ComfyUI 0.18.5 and the workflow templates are in the Template Library's Video section. Key caveat: this is API-access pricing, not open weights — open-source Wan stops at 2.2. Anything calling itself "Wan 2.7 open weights" is a fake checkpoint.
Horror and true crime faceless channels are running $8-13 RPM with strong watch time and search-demand floors that hold up year-round. The format is AI-narrated Reddit-style storytelling over moody generated visuals — fully pipeline-able. Nexlev's July 2026 niche data puts these in the top tier for monetization speed, and several purpose-built AI platforms have spun up specifically for faceless horror and true crime production in the last six months.
This one has been active for a year but operators are still getting caught by it: YouTube's July 2025 clarification means channels using templated formats, AI voiceovers without substantial transformation, or low-originality repurposed clips risk losing monetization across the entire channel — not just the offending video. You need clear creative differentiation baked into your format, not just a different thumbnail.
ByteDance's Seedance 2.0 dropped in February 2026 and benched ELO 1,269 on text-to-video and 1,351 on image-to-video — beating Kling 3.0, Veo 3, and Runway Gen-4.5 at launch. It went viral fast, got a Disney cease-and-desist within days over character generation, and US availability is still blocked pending IP resolution. No announced launch date. This one matters when it opens up — keep it on the radar.
