Consistent Character with this FREE AI Video Generator | Vidu AI
Howto & Style
Introduction
In this article, we will explore the video generation capabilities of Vidu's new Reference to Video feature. This innovation aims to maintain consistency across characters and objects in generated videos, making it an exciting tool for creators. We will evaluate how well it performs with realistic portraits, animated characters, animals, and even merchandise. Let’s dive in!
Getting Started with Vidu AI
Heading over to V. Studio, sign in to your account. On the V dashboard, you'll find three ways to create videos: Image to Video, Reference to Video, and Text to Video. If you’re already familiar with AI video tools, you may recognize the first two options. Text to Video allows you to type a description, which the system converts into a video, whereas Image to Video uses an image for the video generation. Today, we will specifically focus on the Reference to Video feature.
How Reference to Video Works
To begin with the Reference to Video tab, upload your character's image. For this demonstration, I created a realistic character using Mid Journey. After uploading, you can draw boxes around specific areas of your character to guarantee consistency in those regions. While it’s often unnecessary to specify the whole image, I opted to focus on the hairstyle, face, and parts of the clothing.
Next, a straightforward label is needed to describe your character. In this case, I used "a beautiful Asian woman" for clarity. According to the VU prompt guide, an effective prompt generally includes elements like subject, environment, and artistic style. For simplicity, I structured my prompt based on the what, where, when, and how of the video concept.
For my example:
- What: A beautiful Asian woman reading a book
- Where: At a coffee shop
- When: During golden hour (ideal for photography)
- How: Low-angle shot with depth-of-field effect, utilizing a Sony FX6 camera to create a warm and cozy vibe.
Feel free to enhance your prompt with extra details or leave it to V’s capabilities for automatic optimization. For video duration, options include 4 or 8-second clips, with upgrades necessary for longer durations.
Once everything looks good, hit create. Generating a 4-second clip costs 4 credits, while an 8-second clip costs 8 credits—a consideration since free users receive 80 credits monthly, while paid plans start at $ 7.99.
Video Generation and Results
Upon creating a video, it typically takes about 30 to 60 seconds to generate. The initial resolution is relatively low (688 by 384 pixels), which allows for quicker iteration, but you can upscale to 1080p if you like what you see. I was thoroughly impressed with how consistently the character's facial features, hairstyle, and clothing matched my reference image, even from a different angle.
However, several issues arose: the video quality remained blurry, and the character’s reading situation didn’t appear quite natural. If clarity is paramount, consider processing the videos through additional AI upscaling tools.
By using the re-edit function, you can regenerate or edit prompts for further refinements. The tool maintains character consistency remarkably well, yet occasionally struggles with anatomically correct movements.
Testing Animated Characters and Merchandise
I also experimented with animated characters, maintaining consistency in facial features and clothing details. Notably, the AI demonstrated its ability to include additional elements like butterflies with movement around a character, adding dynamism to the scene.
In testing merchandise, I used a handbag featuring the word "coach," which proved challenging as ai tools often struggle with maintaining text integrity. While Vidu succeeded in replicating the bag's shape, text clarity often suffered. Similar results were obtained with a pink Stanley Cup, where color discrepancies arose, underscoring the tool’s strengths and weaknesses.
Conclusion
The Reference to Video feature by Vidu demonstrates impressive character consistency, especially in animated styles. However, it still has improvement areas, particularly with video quality and text accuracy on merchandise. Overall, this tool showcases significant potential for creating engaging video content.
Keywords
- Vidu AI
- Reference to Video
- Character consistency
- AI video generator
- Merchandise testing
- Video quality
- Animated characters
FAQ
1. What is Vidu AI's Reference to Video feature?
Vidu AI's Reference to Video feature allows users to upload character images and maintain character consistency across generated videos.
2. How does the generation process work in Vidu AI?
Users upload a reference image, specify areas for consistency, and provide a prompt describing the desired scene. The tool then generates a video based on this input.
3. What are the limitations I might encounter using Vidu AI?
Potential limitations include video quality (especially in initial previews), accuracy of merchandise text, and occasionally awkward anatomical movements in generated characters.
4. Can I create longer videos with Vidu AI?
Yes, but longer videos (8 seconds) require a paid plan for generation.
5. How consistent are characters across different scenarios with Vidu AI?
Vidu AI demonstrates impressive consistency in facial features and clothing details across various video scenarios. However, further testing may be required for specific use cases.
