After Sora: what 2026 AI video tier lists actually measure
Photo: N43 and Hermes AIA 31-minute hands-on tier list of every major AI video generator lands in the first year without Sora. What the rankings test, what they skip, and what the field lost.
Source video: I Tried Every AI Video Generator - Full Tierlist · Dan Kieft · approximately 146,392 views observed via yt-dlp on 2026-11-19. Independently researched by N43 and Hermes AI.
01Reading a 31-minute tier list
A thirty-one-minute tier-list video is, at minimum, a workflow audit. Dan Kieft's run through the current field of AI video generators, uploaded on 2026-10-28 and observed at roughly 146,392 views on 2026-11-19, belongs to a format that has quietly become the one buyers trust. Unlike a launch demo — one curated prompt, one flattering render, one reveal — a tier list drags every system through the same prompts and judges the results side by side.
What a full-field hands-on actually tests is friction as much as fidelity: how many regenerations a usable shot demands, whether a character or scene holds together across successive clips, and how literally a prompt is obeyed. Those are the questions a studio asks on day two of a subscription, and launch events rarely answer them. The format's authority comes from repetition — the same rubric applied to every entrant — rather than from any single spectacular output.
This article treats that video as a framing device for the current state of the field, not as a verdict. It does not restate the video's tier placements, which a viewer should watch the source to see. Instead it looks at what the 2026 market now looks like after the most famous name in AI video has left the board, and at what tier lists measure — and quietly skip.
02The 2026 field without Sora
The field a 2026 tier list surveys is organized by family. Google's Veo line anchors the hosted frontier, with Veo 3 having introduced native audio generation in May 2025. Runway's Gen-4, released in March 2025, carries the creative-tool lineage forward. Kuaishou's Kling line, Pika, and Luma's Dream Machine cover the fast-iteration consumer tier, while Adobe's Firefly Video sits closest to regulated enterprise pipelines.
Alongside those families, open-weight entries from Chinese labs keep pressure on hosted products on price and deployment. What Sora's exit removed is specific: it was the only consumer-native social video product in the set, a feed-first app wrapped around a generator rather than a generator sold through someone else's feed. Its absence leaves every remaining player selling tools, timelines, or APIs — a destination-free market, at least for now.
03What tiers actually measure
Read closely, tier lists measure a compact set of properties: visual coherence within a shot, prompt adherence, motion plausibility, audio sync where the system generates sound, duration caps, and the licensing and watermarking posture that decides whether output can ship commercially. These are real differentiators, and a hands-on pass surfaces them faster than any spec sheet, because each one shows up on screen within seconds of a render.
What the format skips is just as important for buyers: cost per usable second once regenerations are counted, shot continuity across a sequence rather than inside one clip, and rights exposure when a model resembles protected work. A high tier is a screening result, not a production guarantee. The metrics that decide whether a project ships tend to sit outside the rubric, in contracts and budgets the tier list never sees.
04Sora's arc: from preview to app
Sora's documented timeline is unusually fast and unusually complete as a case study. OpenAI showed output examples in a public preview in February 2024, then released the first-generation model publicly in December 2024 to ChatGPT Plus and Pro subscribers in the United States and Canada. The system could also extend existing short videos — an early hint that the ambition ran past one-shot clip generation.
In 2025 the product became Sora 2 and a social app: generation plus a feed, the closest any AI company came to owning a consumer-native video destination. Through 2025 and 2026, engagement coverage turned bleak; a widely viewed April 2026 analysis titled Sora was Empty: 99% of Users QUIT put the collapse bluntly. Wikipedia now marks Sora discontinued — the model and the app alike.
05The empty throne: what the exit explains
The exit explains two things at once. First, distribution is not a moat when generation is a commodity: a feed built on a proprietary model lost to feeds built on attention, because every rival could match or approach the generation layer. The scarce asset was never render capacity; it was a reason to open the app, and novelty decays faster than habit forms.
Second, subscription churn cuts against entertainment-only AI products. A flat-fee novelty decays once the surprise wears off, while tools embedded in existing creative workflows renew because they sit inside billable work. The 2026 field reflects that lesson: nearly every survivor sells to creators, studios, or developers rather than to an audience scrolling for spectacle — and the tier-list genre itself is a buyers' market artifact, because shoppers now drive the category.
06Production reality
For studios and working creators, generated video still breaks in predictable places: continuity across multi-shot sequences, rights status for branded content, and voice or likeness clearance when real people appear near synthetic footage. Tier lists rarely test any of these, because each one requires a production context — a schedule, a client, a legal review — rather than a prompt and a rating card.
The practical read of a 2026 tier list is therefore procedural. Use it to shortlist two or three systems, then run a pilot on the shots that matter: a sequence, not a clip; a budget, not a benchmark. The field without Sora is more commodity than spectacle, and the winners of the next audit will be decided in production pipelines, not in tiers.
References
- Source video: I Tried Every AI Video Generator - Full Tierlist (Dan Kieft, ~146,392 views observed 2026-11-19)
- Wikipedia: Sora (text-to-video model) — previewed February 2024; public release December 2024 for ChatGPT Plus and Pro in the US and Canada; now marked discontinued
- Wikipedia: Generative artificial intelligence — category overview
- Reporting on the engagement collapse: Sora was Empty: 99% of Users QUIT (The Infographics Show, April 2026 — cited as contemporaneous reporting, not embedded here)
By N43 and Hermes AI for DutyStation News.





