Sora is here

Original Article ↗ Hacker News Discussion ↗

OpenAI’s release of Sora, a text‑to‑video model now bundled into paid ChatGPT plans, is drawing mixed reactions over its technical capabilities, pricing, and strategic timing. Commenters note that while Sora’s samples are impressive, they still suffer from physics and continuity glitches, and several open-source or Chinese models such as Tencent’s Hunyuan are seen as competitive or better—especially given their fine‑tuning flexibility and lower costs. Alongside concerns about limited access (no EU/UK yet) and a lack of API, many worry Sora will accelerate an already growing wave of AI‑generated “slop” and deepfakes, further eroding trust in online media even as it lowers barriers for advertisers and casual creators.

Overall reception & capabilities

  • Many find Sora impressive as a text‑to‑video demo, especially landscapes, bokeh shots, and temporal consistency vs older models.
  • Others say the quality is “ok but not amazing,” still plagued by object permanence, physics errors, uncanny motion, and morphology glitches.
  • Several note cherry‑picking and that some early showcase clips were post‑processed by VFX teams, so raw output is weaker.

Comparisons & open‑source ecosystem

  • Frequent comparisons to Kling, Hailuo, Runway, Luma and especially open‑source Hunyuan Video, Mochi, and LTX.
  • Some claim Hunyuan already matches or beats Sora on quality and cost; others see Sora slightly ahead on detail and consistency.
  • Strong sentiment that, like DALL‑E vs Stable Diffusion/Flux, open‑source video models will commoditize this and win for serious creators (fine‑tuning, ControlNet, ComfyUI workflows).

Pricing, product, and availability

  • Sora is bundled into ChatGPT Plus and Pro, with credit limits, watermarks on Plus, 720p/5s vs 1080p/20s on Pro; no pay‑per‑use or API yet.
  • Many call the $20 tier effectively a trial and see strong nudging toward the $200 Pro plan.
  • Not available in EU/UK/Switzerland; reasons debated (AI Act, DMA, or just capacity).
  • Launch marred by sign‑up lockouts and “heavy traffic” messages.

UX, prompting, and control

  • Common frustration: huge iteration and prompt engineering needed; tools rarely produce exactly what’s in the creator’s head.
  • Text‑only control is seen as fundamentally too low‑bandwidth; people want storyboards, white‑paper‑style briefs, sketch‑to‑video, keyframe control, in‑painting, and character/scene consistency.
  • Some note Sora’s UI emphasizes editing, trimming, and remixing clips rather than one‑shot perfect generations.

Use cases and industry impact

  • Short‑term: seen as “better stock video,” social‑media slop, ads, explainer content, and prototyping for creatives.
  • Opinions diverge on film/TV: some predict big pressure on VFX and low‑budget content; others insist current models are far from replacing Hollywood craft or long‑form coherent storytelling.

Ethics, safety, and “AI slop”

  • Intense worry about misinformation, deepfakes, harassment (e.g., fake nudes), and the loss of trust in online video.
  • Many complain about already‑rampant AI “slop” on social media, YouTube, and search, and expect Sora‑like tools to accelerate this.
  • Some call for watermarking and legal labeling of AI‑generated video; others argue responsibility should rest on users, not tools, and warn about over‑regulation and censorship.