Can Gen-3 Alpha unseat Sora?
💥As the digital content creation world keeps advancing, the rise of Sora as a leading AI platform is creating a buzz in the industry. It introduces a fresh approach where text is easily turned into captivating AI-generated videos, setting a new standard for productivity and creativity. Since the release of more similar models like Vidu, Luma, and Kling that rival Sora, the AI video landscape has entered a new round of competition. Finally, Runway has stepped into the arena with its new weapon, Gen-3 Alpha.
Hi, welcome to The Briefy Blog. Here we will share Briefy's product updates and our best tips for using AI to improve daily productivity. Subscribe to our blog so you don't miss the up-to-date information.
What has improved in Gen-3 Alpha compared to Gen-2?
After Runway's last significant update to Gen-2 in November 2023, many believed that Runway was falling behind. However, with the release of a stunning video, Runway has reiterated its presence as a top player in space. ✨The Gen-3 Alpha represents the first model in a forthcoming line of models that have been honed using a fresh infrastructure designed for extensive multimodal. Runway is claimed to offer significant enhancements in fidelity, consistency, and motion compared to its predecessor, Gen-2. More than this, Gen-3 Alpha also made crucial improvements in video generation quality, color accuracy, saturation, lighting effects, semantic text restoration, camera movement, action consistency, and scene transitions. In addition to technical breakthroughs, Gen-3 might employ a DiT framework to that of Sora, which allows it to accurately simulate real-world scenarios.🌆
*Text Prompt 1: A bustling cityscape at sunrise, with tall skyscrapers casting longshadows, the sky painted in hues of orange and pink, and early morning traffic on the streets below; *Text Prompt 2: A serene forest scene with a waterfall cascading down moss-covered rocks, runlight filtering through the dense canopy, creating dappled patterns on the water surface
If you generate more sample videos using the same text prompt, whether it's characters or scenery, the biggest problem of Gen-2 generated video is unreal, whereas Gen-3 Alpha has already overcome this obstacle. 📈It has also been greatly improved in terms of the generation speed. Gen-2 takes about two minutes to generate a 4-second video, and users have to wait in line to get the video instantly. However, Gen-3 Alpha takes only about 60 seconds for a 5-second generation and 90 seconds for a 10-second generation.
Feature highlights of Gen-3 Alpha
Video generated by Gen-3 Alpha downloaded from runwayml.com
🕹️ Fine-grained temporal control to generate videos with creative shifts and accurate key-framing of elements in the scene.
👩🦲 Proficient in creating dynamic human figures with a broad spectrum of movements, expressions, and feelings.
👩🎨 Involved efforts from researchers, engineers, and artists from various fields to understand a broad spectrum of artistic styles and film-related terms.
📌 Work closely with top entertainment and media companies to tailor Gen-3 models to more specific needs by providing customization.
How to access Gen-3 Alpha?
Gen-3 Alpha is available to the general public from July 2nd. You must have a paid Runway account and it is going to power all of the Runway tools.
- Sign in to your Runway account -> go to Text/Image to Video
- Select the Gen-3 Alpha model from the model dropdown at the top.
- Enter a prompt and select the duration. You can either choose a 5-second or 10-second generation.
- Click "Generate". Gen-3 Alpha takes 10 credits per second for generation and the resolution is going to be at 720p at 24 frames per second at the current stage.
Gen-3 Alpha vs. Sora
In comparison, while Sora is also a powerful video generation model, its performance seems somewhat inferior when measured against the demonstration effects of Gen-3 Alpha. The demo video of Gen-3 Alpha not only reaches high-definition standards in visual effects, but also showcases a higher level of creativity and technical integration. 🧠 For instance, in a video showcasing the ability of Gen-3 to analyze the semantics of the text, the prompt was "top view timelapse video of a pencil artwork drawn by a hand, it is an art of a girl with rabbit hair from beginning till the finish." Despite the type in the prompt (it should have been "rabbit-eared girl"), the final output generated by Gen-3 was remarkably accurate, highlighting its exceptional color presentation capabilities and detailed expression of complex human emotions.😊
However, Gen-3 does have its shortcomings. 🎵It cannot auto-generate background music so that requires users to add sound to all works. Currently, only Google's VideoFX is capable of producing videos with music. Additionally, its handling of lighting changes is always inaccurate, sometimes resulting in vague character expressions. These shortcomings provide direction for future improvements while also highlighting the areas where Gen-3's technology still has room for enhancement.
The table below fills in the detailed comparison between Gen-3 Alpha and Sora from different perspectives:
| OpenAI: Sora | Runway: Gen-3 Alpha | |
|---|---|---|
| Overview | Creates cinematic videos from text prompts, aiming for realism and a professional look. | Focuses on enhancing detail, consistency, and motion representation in AI video generation. |
| Video quality | Sets a new benchmark in realism, with meticulous detail and seamless blending with real-world footage. | Excels in detail and coherence, delivering visually striking and consistent videos. |
| User experience | - Offers a seamless and intuitive workflow. - Prioritizes user interaction and satisfaction. | - Characterized by user-centric design and functionality. - Offers more extensive tools for creative control. |
| Industry applications | Focuses on achieving unparalleled realism tailored for immersive storytelling applications. | Partners with Getty Images to have access to an extensive collection of top-notch images is essential. |
| Availability | Limited release to select groups of designers, filmmakers, and artists | Open to the general public |
| Pricing model | - Pay-as-you-go model - Depends on the number of tokens and the various OpenAi models used - Multiple models, each with different capabilities and price points - 1,000 tokens is about 750 words | - Free trial: not available for Gen-3 Alpha - Standard: $15/month, $144/year - Pro: $35/month, $336/year - Unlimited: $95/month, $912/year - Enterprise: $125 per editor per month |
Who will be the next top?
💭Runway's release with its highly anticipated Gen-3 Alpha model indeed sets a new standard for AI-generated video content, yet the development of AI video generation technology still faces multiple challenges. On the technical front, enhancing video quality and efficiency, and ensuring content coherence and consistency are critical issues in urgent need of resolution. Moreover, the high demand for computing power and expensive inference costs are significant obstacles limiting its commercialization. Ultimately, whether AI-generated videos are satisfactory still largely depends on the users' preferences. Which one do you think is doing even better, Gen-3 Alpha or Sora?🤔
Seeking a better method to gain knowledge and keep updated with new tech coming out? Choose Briefy now and change the manner you consume excessive information. Briefy provides you with concise summaries in different views that save you time and ensure you're up to date.
👏 Follow us on X, LinkedIn, or subscribe to the Briefy Blog from today!