The AI video landscape has shifted dramatically. As of late October 2026, creators in Spain and across Europe face a market where the dominant question is no longer which app to open, but which model to select inside an aggregator platform. The most notable casualty is OpenAI’s Sora, which shut down its web and mobile apps on April 26, 2026, with its API following on September 24, 2026. What remains is a fragmented but capable ecosystem of tools, each with its own watermarking approach and legal constraints.
- Key Point 1: Sora is discontinued, but Google’s Gemini Omni and Veo 3.1, Magnific, Adobe Firefly and Runway remain active, with varying access tiers and capabilities.
- Key Point 2: Google-generated videos carry invisible SynthID watermarks and C2PA Content Credentials, while detection tools remain partial and provider-specific.
- Key Point 3: Usage policies explicitly prohibit impersonation, non-consensual intimate content, biometric misuse without consent, and falsely claiming AI content is fully human-made.
The Platforms Still Standing
Google’s video ambitions are split across two models. Gemini Omni, available only through paid Google AI plans (Plus, Pro and Ultra) and restricted to users over 18, offers video-to-video editing, chained edits within a single conversation, aspect ratio selection and preset style templates. Users can upload one video and up to five reference images, request native audio generation, and download or share results to YouTube. Each clip takes several minutes to render. The free tier excludes video generation entirely.
Veo 3.1, accessible through Google Flow, remains the company’s flagship video model according to DeepMind, delivering native audio with 1080p and 4K output. Meanwhile, Magnific (formerly Freepik AI Suite) aggregates third-party models including Veo 3.1, Kling 3.0, Seedance and Wan 2.6, letting creators pick different engines per shot. Adobe Firefly combines its own Firefly Video Model with partner models from Google and Runway, while Runway promotes its Gen-4.5 as the world’s best video model, a claim made by the company itself.
The strategic takeaway is clear: tying an entire workflow to a single tool is risky, as Sora’s shutdown demonstrated. Aggregators like Magnific and Firefly offer resilience by letting creators switch models without abandoning their pipeline.
What You Can Actually Do
Modern AI video generators handle text-to-video, image animation, video referencing and increasingly conversational editing. Gemini Omni allows users to upload a clip and request changes, chain edits across a conversation, and reference up to three seconds of video to maintain character consistency. It also generates quick drafts at 360p before upscaling to 1080p or 4K.
However, European users face a notable limitation: uploading your own videos for editing is unavailable in the European Economic Area, the UK, Switzerland and some US states. You can edit clips the model creates, but not your own recordings.
For commercial use, licensing matters as much as quality. Adobe states its Firefly Video Model was trained on licensed content such as Adobe Stock and public domain works, making outputs suitable for commercial use, though partner models may carry different terms. Magnific asserts that generated content belongs to the user, including for commercial work, and that it does not train models on user creations.
Detection, Watermarks and Prohibited Uses
Videos created or edited with Gemini Omni across the Gemini app, Google Flow or YouTube carry two signals: the invisible SynthID watermark and C2PA Content Credentials. SynthID is embedded at creation, does not affect quality, and is designed to survive cropping, filters, frame rate changes and compression. Users can verify a clip by uploading it to Gemini and asking whether Google AI created or modified it. Google has also opened its SynthID Detector portal to everyone, recognizing content from partners including OpenAI, NVIDIA and Kakao.
C2PA is a separate open standard backed by a coalition including Adobe, Google, Meta, Microsoft, OpenAI and the BBC, adding an origin and edit history to files. The C2PA organization compares it to a nutrition label for digital content. The limitation is obvious: SynthID only identifies content from Google and its partners, and a detector finding nothing does not prove a video is real.
Google’s Generative AI Prohibited Use Policy, last updated December 17, 2024, bans impersonating any person, living or dead, without explicit disclosure and with intent to deceive; generating or distributing non-consensual intimate imagery or explicit sexual content; using personal or biometric data such as a face without legally required consent; and fraud, harassment, or claiming AI-generated content is entirely human-made. Educational, documentary, scientific and artistic exceptions apply. Gemini’s help pages add that users must hold rights to uploaded images and that violations can result in account termination.
The practical rule is simple: if a video only works because someone will believe it is real, do not make it. The technology to generate convincing clips is available with any paid subscription, while detection arrives piecemeal and only for some providers. Until that asymmetry closes, responsibility rests with whoever publishes.