Recently, Hedra Labs released a research preview of Character-1, a technology that allows users to generate dynamic videos of individuals speaking and singing based on any person's photo and voice content.
Imagine uploading a photo of a person, adding any voice content, and instantly creating a dynamic video where the person in the photo comes to life, speaking or singing, with lip movements, expressions, and gestures perfectly matching the audio! Isn't that exciting?
Official Demonstration Video by Hedra Labs
Key Features and Highlights:
Multi-platform Compatibility: Whether on desktop or mobile devices, users can easily use Character-1.
Unlimited Duration Generation: The preview version currently supports the generation of 30-second videos. With sufficient H100 supply, 90 seconds of content can be generated every 60 seconds.
Support for Various Expressions: Character-1 not only supports dialogues but also handles singing and rapping.
Hedra offers a user-friendly interface, allowing even non-professionals to quickly get started. Users can visit Hedra's official website, use the text-to-speech feature or upload audio files, input character descriptions, and then generate dynamic videos.
Hedra's AI technology ensures high quality and realism in the video content, with expressions, gestures, and voice synchronization all achieving satisfying results.
From the official cases provided, Character-1 can perfectly perform singing, acting, and portraying different emotional roles. The technology is not limited to human characters but can also generate expressive inanimate objects, as long as they have distinct facial features.
The usage is also very simple, with the following steps:
Open the Hedra experience address: https://top.aibase.com/tool/hedra
Upon entering the page, you can see the operation interface
The interface is straightforward; in the first box, input your character's dialogue and select a voice. Alternatively, you can import your own audio if you prefer not to use the generated audio.
Here, I simply input "Hello, this is a speaking video created by AIbase, today we're experiencing Hedra, making video generation as simple as breathing."
Then, in the second box, upload the photo of the person you want to speak. Here, I upload a beauty portrait I previously created.
If you don't have a ready photo, you can directly input your character in the text box below and click create to generate one.
After uploading the photo, click the generate video button below the third box.
Here is the generated video effect:
It can be seen that the speaking video generated by Hedra is quite lively; not only do the lips move, but other parts of the body also move, and there are expressions. However, due to the limited selection of voices available on the platform, the foreign accent does not match my photo character very well. Another drawback is that the generated video is significantly blurrier than my original photo. I hope the platform can improve the video quality in the future.
Here, I upload an audio file myself. I directly use Jianying to generate the audio, selecting a female voice, and input the text to be read.
Retest:
Select import audio
The generated effect is as follows:
The blurriness issue can be resolved using the video enhancement feature of Krea AI. However, note that the free experience is limited to videos under 10 seconds; longer videos need to be cropped. Also, avoid selecting too high a frame rate; I chose 60 frames per second and had to upgrade to a paid version halfway through, which was quite disappointing.
