It’s a new day, and there is a new AI video generator in town. Hotshot, a new AI video generator, seems to offer a free alternative to some big models like OpenAI Sora, Runway, etc. The text-to-video AI model has already showcased some impressive results in AI video creation. The model is still in previews and has been reportedly trained on over 600 million video clips. The video model can generate video footage up to 10 seconds at 720p based on the user’s preference.
Reportedly, videos generated by Hotshot can be extended, remade in higher resolution, or even linked to audio. Several users who have used the model claim that even without additional tools, one can get a single text prompt to make videos in different styles and even combine animation and live-action footage.
If you are looking forward to trying this AI video model, it is currently available on the Hotshot website for free. The free videos do not have watermarks. Free users will only be able to generate three videos per day. Besides, the prompt field also keeps track of how many videos are left for a user. In case a user tries to generate more videos, the platform displays its paid tiers.
The video generation can be seen once you input the prompt. At first, you will see a large video screen, then it starts to fade into a grainy screen. At the bottom, one can see the progress bar, which gives an idea of how long it will take to create the video. Once the video is generated, users can play it, copy the link, or even download it. The downloaded video is in the MP4 format.
Our experience with Hotshot
We tried the AI video generator. For the first attempt, we gave the prompt ‘blue eyes that zoom into the grand canyon’. Although the output wasn’t something that we envisioned. The skin texture of the woman, eyelashes, eyebrows, and hair were impressive. In fact, the legendary Grand Canyon looked realistic, and the skyline was of a bright, sunny day with a small hint of sunlight on the woman’s nose and cheeks.
For the second video, we used the prompt – ‘A panda surfing on sand’. This video, an unlikely scenario in real life, was something that caught our attention. The model generated the video as per the prompt; however, the footage had a few telltale signs suggesting that it was an output of an AI model. For instance, the size and shape of the surfboard are inconsistent. Similarly, the panda seems to have three limbs.
The third video was generated using the prompt: Corgi dog driving a car. This was the most visually appealing and accurate creation by Hotshot. The canine looked lifelike with its small paws on the steering wheel. The details of the inside of the car seemed pretty accurate. Overall, the video was clean and good to look at.
Below are some Hotshot samples from X:
Faces created by @HotshotSupport ⬇️🧵 pic.twitter.com/NJJKXgmqeO
— توليف (@tawleefai) August 25, 2024
Wide angle slowly transitioning to close-up: A vast field of sunflowers stretches to the horizon under a blue sky. The camera slowly zooms in, revealing a Brazilian woman lying among the flowers. Her expression is serene and dreamy. Soft focus on the background.… pic.twitter.com/b2W5gg29oH
— Aakash (@aakashsastry) August 24, 2024
#hotshot a diver in clear blue ocean, surrounded by aquatic life forms swims to the camera and holds up a whiteboard with the word “love” written on it pic.twitter.com/lIo9mtEoyE
— Kinky Kong.AI (@Kong_Kinky1) August 22, 2024
See the map in the bottom left 🙂
A shaky handheld video of a first-person shooter chase through the forest pic.twitter.com/tYI2rbpBq7
— Aakash (@aakashsastry) August 21, 2024
Try it yourself
To try the model, one will have to log on to hotshot.co using a legitimate email address. On the home page, one can see the videos created by them and some featured videos generated by the model. There is also a prompt guide to help users frame their prompts better.
Hotshot is a startup that was founded in 2023 by Aakash Sastry, Duncan Crawbuck, and John Mullan. The platform made waves last year as a free-to-use AI photo creation and editing app. Reportedly, the AI video model was trained in just four months by a team of four engineers.
