RCTV.
W/28 · 2026Stack
ISSUE 28 A publication on AI video.Written for people who already know what Sora is. JUL 13 · 2026
Roundup · Weekly

AI Video Weekly Roundup — July 13, 2026

Kuaishou's Kling closes a ~$3B round at an $18B valuation, the largest ever for a video-AI firm; five vendors ship assembly-layer products; Meta previews Muse Video, already #3 on Design Arena.

JUL 13 9 MIN ROUNDUP Read the piece
Living reference The AI Video Stack, 2026 Every model, tool, and service worth knowing — tiered by what they do, not how loud they are. Updated weekly. Open Stack

This week, further down

3 pieces
Mode A analysis infographic, 'The Model Isn't the Moat. The Production Is.': six tiles arranged 3×2. Card 1 — '$500' Sara Nambiar spent over two weeks making videos with HyperFrames plus Codex before concluding the hard part was production, not generation. Card 2 — '$100' Anthropic engineer Thariq Shihipar's Fable usage to edit Anthropic's own launch video. Card 3 (accent) — '6+' production cost centers (timing, audio sync, continuity, captions, QA, assembly) that sit around scene generation. Card 4 — '9' reusable production skills in Nambiar's open-source repo; nobody systematizes 'generate a good scene.' Card 5 — '0' traditional video editors opened; Thariq directed the agent through FFmpeg, Remotion, and the Figma MCP. Card 6 (accent) — 'n=2' field reports plus one artifact: two creators, two toolchains, the same conclusion — early but convergent.
Flagship · Analysis

The Model Isn't the Moat. The Production Is.

Two people, two toolchains, one conclusion in the same month: generating the scene was never the hard part. The moat is everything after it — timing, continuity, captions, assembly, and iterating cheaply toward a target.

JUL 09 · 8 MIN
Six stats from the workflow-layer week: Google's Gemini Omni Flash preview reached Higgsfield's MCP in 38 minutes; Omni Flash's real spec is $0.10 per second with a 10-second cap in public preview; Omni Flash debuted at Elo 1,404 (#1 on Design Arena); HeyGen's HyperFrames launch achieved 1.53 million impressions; Higgsfield's MCP price dropped to $0; and Seedance 2.5's native duration is 30 seconds.
Roundup · Weekly

AI Video Weekly Roundup — July 6, 2026

Gemini Omni Flash hit public preview and was inside Higgsfield's MCP in 38 minutes; HeyGen, Higgsfield, and Runway fight it out at the workflow layer; Idaho's law goes live as Arizona vetoes AI bills.

JUL 06 · 8 MIN
Six stats from the convergence week: Seedance 2.0 native 4K on three platforms in 24 hours, 387K impressions on Higgsfield's 4K launch, Google's reported $75M investment in A24, Seedance 2.5 previewed at Volcano Engine FORCE with 30-second native clips, HappyHorse-1.1 debuting at Elo 1,153, and Idaho's synthetic-media voyeurism law effective July 1.
Roundup · Weekly

AI Video Weekly Roundup — June 29, 2026

ByteDance previewed Seedance 2.5 at FORCE and confirmed 4K for 2.0; three platforms distributed it in 24 hours; Runway declared itself an aggregator; Google DeepMind invested in A24; HappyHorse-1.1 debuted at #2.

JUN 29 · 9 MIN
Living reference

The AI Video Stack, 2026.

The most current comparison of AI video software in 2026 — Veo 3.1, Kling 3.0, Seedance 2.0, Grok Imagine 1.5, Luma Ray 3.2, Runway, HappyHorse and more. Updated every week as models launch and pricing changes.

Last update · Yesterday Open Stack →

Recent, in order

View full archive →
JUN 22
ROUNDUP
AI Video Weekly Roundup — June 22, 2026
Read →
JUN 15
ROUNDUP
AI Video Weekly Roundup — June 15, 2026
Read →
JUN 11
FLAGSHIP
The Model Was the Product. Now the Agent Is.
Read →
JUN 08
ROUNDUP
AI Video Weekly Roundup — June 8, 2026
Read →
JUN 04
ANALYSIS
Grok Imagine vs. Seedance 2.0: the Wrong Question
Read →
JUN 01
ROUNDUP
AI Video Weekly Roundup — June 1, 2026
Read →
Mondays, before 9am PT
The week in AI video, without the hype tax.