RCTV.
W/46 · 2026Stack
ISSUE 46 A publication on AI video.Written for people who already know what Sora is. NOV 16 · 2026
Roundup · Weekly · Nov 16, 2026

AI Video Weekly Roundup — November 16, 2026

Gemini Omni 1.2 Flash tops Design Arena but lands mid-pack on Artificial Analysis. Runway, fal and Luma move into the chat apps, and Seedance 2.5 reaches 4K at Runway's price.

Infographic of six numbers for the week ending November 13, 2026: Gemini Omni 1.2 Flash at 1369 on Design Arena's video board; its Artificial Analysis rank range of 3 to 9 on text-to-video with audio; $6.26 against $9.12 per minute for Omni 1.2 Flash and 1.1 on AA; Wan 3.0 at 1196 leading AA video editing outright with range 1 to 1; 110 credits a second for Seedance 2.5 at 4K on Runway Dev; and December 3, when Omni 1.1 Flash requests start billing at 1.2 rates.
AI-GENERATED NOV 16, 2026 7 MIN READ

No lab shipped a new flagship video model this week. The news was grades and shelf space: two leaderboards rated Google’s cheap Gemini Omni 1.2 Flash and reached different verdicts, three video vendors moved into the chat apps’ plugin directories, and resellers put a 4K price on Seedance 2.5, a resolution ByteDance’s own API doesn’t sell.

Models covered: Gemini Omni 1.2 Flash · Wan 3.0 · Utopai X · Seedance 2.5 · FLUX 3 Max · Runway · fal · Luma


📊 Gemini Omni 1.2 Flash is first on one leaderboard and mid-pack on another

Gemini Omni 1.2 Flash went generally available on Nov 5 at about 7 cents a second. This week it got scores, and two instruments disagree about it.

On Design Arena’s video board, read Nov 15, it is first at 1369 across 4,201 battles. Omni 1.1 Flash is second at 1357 and the original Omni Flash third at 1338, so Google holds the top three. HeyGen Video 1.0 is fourth at 1329 and Wan 3.0 fifth at 1318. Design Arena rates by pairwise votes on its own scale and publishes no rank ranges. That scale is not Artificial Analysis’s, and the numbers can’t be set side by side. This is Design Arena’s main video board; its image-to-video and editing boards are separate.

Artificial Analysis is less impressed. On text-to-video with audio (AA-Video-T2V v2.1, 34 models, read Nov 15), Omni 1.2 Flash is fifth at 1152 with a rank range of 3–9. The top is a two-way tie: Wan 3.0 (1175, range 1–2) and Utopai X (1172, range 1–3). Every row from third to eighth has a range that overlaps Omni’s. Without audio it is seventh at 1113, range 6–11, below a five-way tie for first among HeyGen Video 1, MiniMax H3 Max, FLUX 3 Max, Utopai X and Wan 3.0. On image-to-video with audio (v2.1, 35 models) it is fourteenth at 1085, range 14–14, behind Seedance 2.5’s 1161.

Editing is its best result. On AA-Video-Editing v1.1 with audio, Omni 1.2 Flash is second at 1164 (range 2–3). Wan 3.0 leads outright at 1196 (range 1–1). Without audio, Omni 1.1 Flash (1167, range 1–2) and Wan 3.0 (range 1–3) still share first, and 1.2 has no row there yet.

The price column is the other half. AA lists Omni 1.2 Flash at $6.26 a minute against $9.12 for 1.1 on the text-to-video board, about 31% less. The editing board prices 1.1 at $9.64, so name the board when quoting it. $6.26 is Google’s own 1080p rate: 8,688 tokens a second at $12 per million, which Google rounds to $0.104 a second. Across the text-to-video rows from third to eighth, AA’s prices run from $2.40 (HeyGen Video 1) to $34.12 (Seedance 2.5).

Why it matters: Neither board supports a “best video model” sentence. Design Arena’s voters put the cheap model first. AA’s own ranges put it somewhere in the top nine on text-to-video and fourteenth on image-to-video. A buyer who cares about editing has the better evidence.


💬 Video generators move into the chat apps’ plugin directories

OpenAI’s developer account showed the Runway plugin at work in Codex and ChatGPT on Nov 10. Runway’s MCP page says “Now in the ChatGPT app directory.” The connector itself is older; the directory listing is the news. fal followed on Nov 11, and OpenAI’s own fal page lists image, video, audio and 3D generation, 15 bundled skills, and installs in ChatGPT Work or Codex. Luma said Nov 10 its MCP is live in ChatGPT and Claude, and its page offers one-click adds for both. Runway’s page lists Claude.ai and Claude Desktop among its hosts too.

All three bill the vendor’s way. Runway says its connector uses Runway credits and needs no API key. Luma’s page says credits are what each generation draws from a Luma plan. fal’s OpenAI page says it “requires a connected fal account.” None of the pages states a price for the plugin, and none says what the directory costs the vendor.

Why it matters: The chat window is becoming the storefront, and the vendor still keeps the account and the meter. Whether placement carries terms is the question none of these pages answers.


🎞️ Seedance 2.5 reaches 4K at Runway and Pika, though ByteDance’s own API doesn’t offer it

Runway’s Dev API changelog lists “Seedance 2.5 4K” on Nov 11, for text-to-video, image-to-video and reference video-to-video. Its price table bills Seedance 2.5 at 20, 30, 68 and 110 credits a second at 480p, 720p, 1080p and 4K, with credits sold at $0.01. That is $0.20 to $1.10 a second, and a 5-second 4K clip is $5.50 by Runway’s own example. Pika said the same day that 4K is native on Pika and its API Club.

ByteDance’s own API doesn’t sell it. BytePlus’s ModelArk price table, read Nov 15, prices Seedance 2.5 at 480p, 720p and 1080p and marks 4K “Not supported.” Pika calls its 4K native; neither host says how it gets there, and ByteDance has said nothing about 4K for hosts. A faster Seedance 2.5 Flash is also circulating. Dreamina’s own page describes it as a 480p and 720p preview tier with no date or price. Higgsfield says it is nearly twice as fast at lower cost, in early access and not yet in the US. Runway said late on Nov 12 that Flash is available on Runway; its API changelog has no Flash entry.

Why it matters: The only Seedance 2.5 4K price list is a reseller’s, at about 1.6 times the 1080p rate, for a resolution the model’s maker doesn’t sell on its own API.


🗞️ Also this week

  • FLUX 3 Max’s fal promotion. fal’s page lists 50% off at $0.085 a second at 720p against a $0.17 list, “ends November 21.” fal’s model page says the rate runs through Nov 22, adds $0.145 at 1080p, and returns the list ($0.17 and $0.29) on Nov 23 at 00:00 UTC. fal’s own post says Nov 22. Both pages still call it tied for first in image-to-video, from an Oct 18 snapshot of an earlier AA version. On the current board it is sixteenth, range 15–16.
  • Luma Character Swap & Insert is live in Luma, per Luma’s Nov 11 post.
  • Runway Sightline, for 360 panoramas and camera framing, is app-side and not in the Dev changelog.

📈 By the Numbers

  • 1369 — Gemini Omni 1.2 Flash on Design Arena’s video board, first by 12 points over Omni 1.1 Flash (1357), 4,201 battles, read Nov 15. Design Arena’s scale, not AA’s.
  • 3–9 — Omni 1.2 Flash’s rank range on AA’s text-to-video board with audio (v2.1, 1152, Nov 15). Wan 3.0 and Utopai X tie for first.
  • $6.26 vs $9.12 a minute — AA’s price for Omni 1.2 Flash against 1.1 on that board. The editing board prices 1.1 at $9.64.
  • 1196 (range 1–1) — Wan 3.0 on AA’s video-editing board with audio (v1.1), leading outright. Omni 1.2 Flash is second at 1164 (2–3).
  • 110 credits a second — Seedance 2.5 at 4K on Runway Dev, $1.10 at $0.01 a credit. The 1080p rate is 68.
  • Dec 3 — the date requests to Gemini Omni 1.1 Flash start billing at 1.2 Flash rates, per Google’s changelog, 17 days after publish.

🔮 What to Watch Next Week

  • Gemini Omni 1.2 Flash: Design Arena’s image-to-video and editing boards, and a no-audio editing row on AA. Requests to 1.1 move to 1.2’s rates on Dec 3.
  • Seedance 2.5: how hosts produce a 4K that ByteDance’s own API lists as not supported, Flash pricing from ByteDance, and whether Runway’s API lists Flash.
  • Plugin directories: a price or usage terms on any video plugin, and an Anthropic-side directory listing.
  • Kling 4.0 and Flash: Kling’s pricing page is unchanged (Flash 0.6 units a second at 720p), no AA board has a Kling 4.0 or Flash row, and Kling still hasn’t said whether Flash outputs audio. Kuaishou’s Q3 results call is Nov 19, 7:00 PM HKT.
  • FLUX 3 Max: fal’s 50% rate ends Nov 21 or 22, depending on the page, and the $0.17 list returns Nov 23.
  • EU AI Act, Dec 2: the Article 50(2) marking duty for systems already on the market and the new Article 5 ban on non-consensual intimate imagery, both under Regulation (EU) 2026/1744, are 16 days out.

For full specs, pricing, and access details on every model covered this week, see the AI Video Stack 2026 reference page — updated every Monday.

See also

AI Video Stack — 2026

Living Reference · Nov 15
Mondays, before 9am PT
The week in AI video, without the hype tax.