Google has announced Gemini Omni 1.2 Flash, its AI-powered video generation model. The model produces videos of up to 15 seconds, allows editing by voice, and is positioned roughly 30 percent cheaper than the previous version. It may look like a one-line product announcement, but the real story isn't in the technical specs — it's in the price.
What happened?
The new model highlights three things: short but controlled video, editing via voice commands, and output that stays more faithful to reference images. In Google's product language, the "Flash" suffix means speed and low cost — in other words, this is not the most powerful version but the most economical one. Capping generation at 15 seconds fits that same framing.
Why does price matter so much?
AI video generation is an expensive business. Every second of video means a serious GPU or TPU bill behind the scenes. While OpenAI's Sora chases a "creative studio" perception with longer, more cinematic output, Google is playing in a different space here: driving costs down and ramping volume up.
This choice is no accident. Google runs its models largely on TPU hardware it designed itself. While a significant share of competitors depend on third-party GPU supply, a company that makes its own silicon has far more room to maneuver on unit cost. The 30 percent cut is better read not as "generosity" but as the natural result of vertical integration.
The time limit is part of the same logic. Fifteen seconds suits not long-form creative storytelling but high-volume, repetitive production — ads, product promos, social media clips, e-commerce visuals. Here Google appears to be targeting not the artist, but the marketing teams and developers who need to churn out hundreds of short videos a day.
The per-second cost war
Over the past two years, the center of the AI race has shifted from the "biggest model" debate to the "cheapest model to run" debate. We already lived through this on the text side: prices came down steadily, and cost per token became the main axis of competition. The same curve is now starting in video generation.
| Approach | Key value | Typical use |
|---|---|---|
| Long + cinematic model | Creative quality, narrative | Film, concept, production |
| Short + cheap model (Flash style) | Unit cost, volume | Advertising, e-commerce, social media |
What splits this table in two isn't quality — it's the economic model. In the premium, higher-priced segment, visual impact sells the work; in the cheap, high-volume segment, unit cost drives it. Google is aggressively claiming the second column. When one company breaks price, rivals generally have to follow; this cut is therefore likely to kick off price pressure across the sector.
Editing by voice: a quiet but important shift
The overlooked part of the announcement is "video editing by voice." Until now, AI video generation has largely been a one-shot affair: you wrote a long prompt, and if you didn't like the output, you started over. Dialogue-based editing changes that — with sequential commands like "shorten that scene" or "change the background," it becomes possible to sculpt the output step by step.
This carries the transformation seen in text models over to the video side. Moving from the one-shot command to continuous dialogue turns the tool from a "generate button" into something closer to an "editing partner." This kind of iterative control is exactly what generative AI needs in order to enter real workflows.
What's next?
There are a few things to watch. First, the rivals' price response: if Sora and other video models answer this cut, per-second cost could fall fast. Second, whether the time limit is a genuine constraint or a deliberate, price-driven choice — if later versions deliver longer output that's just as cheap, the equation changes. Third, how reliably voice editing actually works in practice; features that look good in demos often stumble in real workflows.
Ultimately, Gemini Omni 1.2 Flash is less a technical breakthrough than a strategic positioning. Google is choosing to compete on economics rather than quality, trying to turn AI video generation from a showpiece into commercial infrastructure. The value of this move lies not in how impressive the video the model produces is, but in how cheaply it produces that video.
Yorumlar (0)
Henüz yorum yok. İlk yorumu sen yaz.
Yorum yapmak için Sinyal'i indir
Yorumlar Sinyal hesabıyla yapılır. Mobil uygulamada giriş yap, yorum bırak.