Viva AI Video Generator: Experience Free AI Video Generation

Video is the jewel in the AI generation technology field and one of the biggest cakes in AI commercialization scenarios.

Since the release of Sora, discussions about video models have been fervent. The long video generation model solution, represented by the DIT technology architecture, has become a new hot topic. After seeing the Vidu model released by a domestic startup and the Veo model unveiled at the Google I/O conference, a recent AI generation model called VIVA, featuring a vertical screen Sora, has opened for free testing!

Let's follow along and explore the power of the VIVA AI video generator!

Highlighting Viva AI

Viva AI is currently the only video generation model similar to Sora that is open for free testing. Within two days of its launch, overseas users had created over 10,000 videos. It is currently open to global users with test ports on its website and Discord. Anyone eager to see the new capabilities of video models can dive in!

From official promotions, VIVA AI shows excellent capabilities in prompt understanding, video continuity, simulating the real physical world, and imagination, producing stunning effects comparable to Sora with the same prompt commands.

Today, I'll reveal the must-see highlights of the VIVA video generation product!

Portrait Mode of Sora

The first highlight you can't miss is the vertical screen video. Most video generation models on the market currently display in landscape mode, but VIVA AI might be the first vertical video generator.

Vertical videos better fit the usage habits of today's main video consumption scenarios, whether it's TikTok, Instagram, or short drama and live streaming platforms. Viva AI likely aims to let more users generate real AI vertical videos with simple inputs, ready to be shared on social media platforms like TikTok and Instagram with music edits.

User-Friendly Operation

Currently, Viva AI video generator is fully open for free testing, which is quite generous and bold compared to other video generation models. Upon entering the website, I found that VIVA has done a lot of work in terms of user-friendly operation.

For most ordinary users, the biggest difficulty in AI generation is writing prompts. Beyond the video model's capabilities, Viva AI offers a Magic Prompt function that transforms your input into video description language prompts, making details like perspective, shooting style, camera movement, and object details more specific for better video generation results.

Strong Generation Control

Looking at the evolution of image generation products, we see a shift from simple prompt-controlled generation to more detailed control abilities with positive and negative terms and even more refined pose and action controls with ControlNet. Users want generation tools with increasing control capabilities. Viva is user-friendly in this aspect by offering several control functions.

In the prompt window, you can edit negative terms and adjust the video generation size. VIVA also supports adjusting the motion strength of video frames. For example, you can precisely control the movement of a green chameleon.

VIVA also offers a 4K enhancement feature, allowing users to enhance the clarity and detail of selected video segments.

Currently, Viva AI is in a free phase on its website. During my testing, I generated many video clips without restrictions. In the public version, each video is limited to 5 seconds, but I found a 15-second video Easter egg on Viva's website, suggesting the product might have longer video generation capabilities. The generation speed is quite fast, supporting up to three concurrent requests.

Comparison with Top Models

Since VIVA claims to challenge the latest video generation solutions like Sora, let's compare its horizontal/vertical video generation with OpenAI's Sora and Google's latest Veo video model.

VIVA AI vs. Sora

Prompt: Photorealistic closeup video of two pirate ships battling each other as they sail inside a cup of coffee.

Prompt: A young man at his 20s is sitting on a piece of cloud in the sky, reading a book.

VIVA AI vs. Veo

Prompt: An aerial shot of a lighthouse standing tall on a rocky cliff, its beacon cutting through the early dawn, waves crashing against the rocks below.

Prompt: Extreme close-up of chicken and green pepper kebabs grilling on a barbeque with flames. Shallow focus and light smoke. Vivid colors.

Viva AI performed quite well using the same test prompts, with an impressive understanding of semantics and smooth visuals. Its vertical videos showed excellent coordination in perspective and composition, indicating possible optimizations for vertical content in its training data and architecture.

However, Viva still has room for improvement. In some aspects, it lags behind Sora in demos.

Sora's demos generally achieve seamless transitions between different camera angles, generate complex scenes with multiple characters and specific actions, and maintain stability during object movement, areas where Viva can still improve.

Viva AI Makes Video Creation Cooler

Many current AI tools differ significantly from previous product tools, as they heavily rely on users' ability to train the AI. The outcome variance in AI tool usage is much larger.

After deeply testing VIVA, I believe its capabilities have reached a basic level, with significant potential in certain video creation scenarios. Since VIVA is currently free for all users, those needing video materials and AI enthusiasts can start exploring!

Scene/Scenery Shot

Vvia AI can now achieve excellent output for stock footage and scene videos. For example, scenes of plants growing rapidly, rivers under the sunset, or star trails moving in the night sky can be stitched together to create a realistic travel documentary.

Vertical videos also show impressive effects, with highly realistic scenes such as swaying flower fields, tranquil underwater views, or rainy city streets. The material tested has achieved a 90+ score in detail, with some scenes difficult to distinguish from real footage.

These types of stock and scene videos have many real-world applications, such as promotional videos, MV backgrounds, vlogger transition clips, and even movie short film segments.

Before AI video generation tools, users needed to search and download paid video libraries or shoot footage themselves. Now, some of these needs can be met with VIVA AI.

Additionally, imaginative scenes that previously required highly complex animation skills can now be generated by simply inserting imaginative descriptions into the prompt, bringing new possibilities to video creation.

Physical Exhibition Scene

In all the tested content, single-task and item showcases showed no significant jitter or flaws, making them very suitable for video advertisements.

For example, a brand advertisement for seasonings can use AI to generate a series of food presentation scenes. Similarly, a sunglasses ad can show various celebrity animals wearing sunglasses in the introductory part.

This type of scene testing is highly operable, providing new creative elements and possibilities for showcase videos. VIVA's rich detail in single-subject videos makes each moment vivid and highly watchable.

Sci-Fi Narrative Scene

The most anticipated aspect of AI video generation compared to previous creative ecosystems is its performance in imaginative scenes, generating stylized content faster.

For example, scenes like a boat in a painting studio, a panda in a fish tank, or a cyberpunk robot in the city can all be realized in AI-generated videos.

Moreover, Viva AI understands and displays interactions between multiple characters relatively well in sci-fi narrative videos.

Sci-fi narrative videos can be used in animation, feature films, short dramas, and similar applications. Professional users can create a storyline with video storyboards.

Cinematic Quality

For cinematic quality scenes involving real person interactions, Viva AI video generator can produce good content, but there is still some jitter in hand details and fast-moving characters, requiring further iterations. However, the overall effect is generally usable.

I tried creating a short film with a storyboard concept, depicting three scenes that combined into an AI documentary about a group of people climbing Mount Everest and eventually reaching the summit.

Scene 1: Preparation Phase

Prompt: A determined climber stands at the base of Mount Everest, meticulously checking his gear for the ascent. The majestic Himalayas tower behind him, with clear blue skies and the sun casting a golden glow on the snow-capped peaks.

Scene 2: Challenge Phase

Prompt: As the camera zooms in, we witness the climber's struggle amidst the harsh conditions of Mount Everest. He navigates treacherous ice fields, each step a monumental effort.

Scene 3: Summit Success

Prompt: The climax of the video shows the climber, alongside his team, finally standing atop Mount Everest. They wave their flags in triumph, faces beaming with joy.

These are my comprehensive findings from deeply testing Viva AI video generator. While it still has some issues, such as video length, semantic understanding and precise execution, multi-angle video generation, multi-character interaction, and scene stability, the VIVA team plans further iterations.

The VIVA AI team recently announced a creator recruitment plan to attract talented creators to generate more exciting content using the tool. The product team also plans to release a version with longer video generation capabilities, inviting creators from the recruitment plan for internal experiences first.