Google reportedly tests Gemini Omni video editing with chat remix and templates

• chrisfirst's UI screenshot shows a Gemini app prompt for "Create with Gemini Omni," with Google-described actions that include remixing videos, editing directly in chat, and using templates.
• In chrisfirst's first-look thread, the standout claim is text fidelity: he says Gemini Omni is "better at text rendering than basically all other video models currently," and his attached demo focuses on editable on-screen text.
• The chalkboard proof clip spread through minchoi's post and ai_artworkgen's reposted leak, both using the same prompt about a professor explaining trigonometric identities while writing on a board.
• The evidence is still unofficial: ozansihay's post frames the clip as something "said to be made" with Google's new Omni model, while LinusEkenstam's commentary post treats it as a broader video-model breakthrough rather than a confirmed launch.

You can already see the product shape in the Gemini UI screenshot: a mobile prompt to create with Gemini Omni, plus verbs for remixing, chat editing, and templates. Then chrisfirst's follow-up demo pushes on a more interesting frontier for creative tools, readable text inside motion, while minchoi's chalkboard clip and ai_artworkgen's repost show why that matters for educational and explainer-style video.

The clearest product evidence is the surfaced Gemini interface, not the flashy demo clip.

The screenshot shows a mobile Gemini prompt labeled "Create with Gemini Omni" and lists three actions:

• Remix your videos
• Edit directly in chat
• Try a template

chrisfirst's first-look thread repeats the same three verbs in text, which makes this look less like a one-off generation endpoint and more like a conversational editing layer inside Gemini.

Video models still fall apart when a scene needs crisp, changing words. That is why chrisfirst's first-look thread zeroes in on text rendering first.

His demo centers on a sentence being typed and revised inside chat while the generated video updates with cleaner lettering. The claim is narrow but useful: if the clip is representative, Gemini Omni is being tested on one of the ugliest failure modes in AI video, live text that stays legible after edits.

The most widely shared example is a professor writing out a trigonometry proof on a chalkboard while apparently narrating the current step.

Both posts use essentially the same prompt, and both attach a clip where the board remains readable enough to sell the scene. For AI filmmakers and explainer creators, that is the part worth clocking. The hard shot is not just "person talking," it is sustained hand motion, symbolic notation, and a surface full of text that viewers actually need to follow.

The leak is traveling through creator and commentator accounts, not through a public Google announcement in this evidence set.

That trail matters because the framing shifts with each repost. ozansihay's post presents the clip as something reportedly made with Google's new Omni video model. LinusEkenstam's post strips it down to "new video model breakthrough." ai_artworkgen's retweet context ties the same chalkboard demo back into a feed otherwise focused on Seedance 2.0 work, which is a good reminder that creators are comparing this rumored Google system against the current crop of production video tools in real time.