A Guide to Content Generation
Explore the nuanced process of video generation on Deepshot through our comprehensive guide. Gain insights into the importance of facial focus and much more.
Hey all,
Feeling lost in the mystical land of video generation on Deepshot? Fret not, my fellow film wizards, we've got a treasure map that will guide you through the roughest terrains of our platform, answering those pesky support questions we get in our magic crystal (or, you know, our support inbox).
Creating Profiles
In the world of Deepshot, the face is the star of the show. And like any good film star, it needs its moment in the spotlight for the full duration of the video. Special emphasis on the lips—they're the real divas here. If they or the face are out of spotlight even for a microsecond—say, for instance, someone coughs, a microphone does a cameo, or the speaker gives us their best dramatic turn away from the camera—be ready to hear "Cut!". This, my friends, is a recipe for a failed video generation. We know this sounds like a strict director's rule, but we assure you, our face detection and HD generation model is about to make its debut soon. This will mean your source video will only need to include face time.
If you are someone that plans to get your video samples from YouTube, I would recommend using a tool like Cut YT, which allows you to paste in a YouTube link and download only a segement of that video.
Too long dont care:
  • Ensure the face is clearly visible throughout the video sample. Special emphasis should be on the lips.
How long does Generation take ?
Wondering how long you'll have to twiddle your thumbs while your video is in the makeup chair? A good rule of thumb is that it shouldn't take more than twice as long as your final video's runtime. However, if there's a red carpet event (a surge in requests), your video might take slightly longer. Patience, dear fans, good things come to those who wait!
Too long dont care:
  • The generation time usually takes approximately twice as long as the video's runtime.
Output Resolution
Lovers of the 9:16 videos, we hear you! We know you want your videos to shine on the social media stage. Currently, our beta version of Deepshot requires you to do a quick costume change of your videos to 16:9 before uploading, and then back to 9:16 post-generation. We know that's like asking a film star to change outfits mid-scene, and we assure you, our next big update is going to save you from this.
Stay tuned for more updates and keep those cameras rolling!
Too long dont care:
  • For the beta version, change your videos to a 16:9 aspect ratio before uploading, and back to 9:16 after generation.
Audio Samples
First off, your audio files should be more 'solo artist' and less 'crowded concert'. This means we need your files to be in .mp3 format and contain only the speaker talking, with as little background noise or music as possible. Imagine you're recording a whisper in a library - that's the level of quiet we're aiming for!
Next up, size matters! We're not asking for a symphony - a short and sweet audio file of 10MB or less is all we need. And length? We're all about the tight, focused sound bites here. No more than 5 minutes of high-quality audio is needed.
If you find yourself struggling to find audio without background noise, I would recommend checking out VocalRemover, a tool that allows you to extract vocals from noisy environments.
Too long dont care:
  • Audio files should be in .mp3 format, contain only the speaker talking, and have minimal background noise.
  • Audio files should be 10MB or less, and no longer than 5 minutes.