How AI can be used in history docs today
The Big Wide #001
AI is already in use, and we think history is the genre it will start to transform first. It offers a number of opportunities for telling new stories:
Ideas that require extensive drama reconstruction or VFX (Ancient Egypt, the Stone Age) can be bought to life by a single person with image to video or text to video AI.
Characters from the recent past can voice their own material (diaries, books).
Poor quality archive can be upscaled and cleaned up at a very low cost.
Niche subjects with little available archive can escape talking heads and be bought to life with images or video.
Here’s five tools that already exist that can make these ideas happen:
Upscale archive stills using Magnific. This tool can massively upres pixelated or old photos. We’ve tried it out, and while there are some slight changes and additions, once you get the hang of the settings results can become reasonably true to the original. Use it to add detail and definition when all you can find are small or old still images, or to zoom in to historic photos.
Upscale and denoise poor quality video with Topaz Video AI 4. This is reported to work especially well for footage from the 70’s onwards.
Bring source material to life with Eleven Labs voice cloning. Want Churchill to voice his own diaries? Upload a sample, wait a few minutes, and you are ready to go - feed in text from his diary and receive back an audio track.
Eleven Labs side note - this also works for reading in temporary narration. Use the professional voice cloning toolto have your own voice read an entire script in minutes.
Need stock archive imagery to illustrate an idea? Feed Midjourney prompts that detail exactly what you are looking for (1930’s house, suburban America, grey walls, drone mid-shot, daytime, Netflix style) and, after a couple of refinements, you’ll have a cinematic quality stills ready to go. Also a great tool for pitch decks and style notesthat we’ve used for over a year now without complaint. Next up - image to video, where we veer into the world of magic 🪄
Bring text or images to life with platforms like Runway ML or Pika Labs. It’s a boom time for AI video and there’s dozens of different platforms, but Runway is our reccommendation. Describe a scene or upload an image to start. Then define camera movement using the arrows and guides (less is more here), and let it get to work. It normally takes a few tries to output usable footage, but when it does it feels quite special. AI video is still in its infancy, but we’ve already seen these tools come on massively in just a few months of using them. We predict they’ll be competing with stock footage fidelity by the end of 2024.
Have a look at this video below - all imagery created entirely by AI with stock music and sound. If this is possible in early 2024, who knows what will be possible in a few years…
The Big Think:
Staying true to reference material will be key as this becomes mainstream. Just like drama reconstruction aims to utilise everything we know about an event, we should be able to reference back all details in the video we create. Proper research is where high quality films can still differentiate themselves.
Reconstruction of historic events may become more true to life - generative AI offers near limitless iterations and changes for a fraction of the cost of a drama shoot.
Eras that haven’t had their time in the spotlight due to perceived lack of interest should brace themselves - the barrier to telling their stories is about to get much lower. Stone Age Britain anyone?
Have a look at the results you can achieve in the video below (where all imagery was generated with AI) or head over to the Curious Refuge channel that keeps tabs on all things video and generative AI.
Film that looks like it was shot last week belongs only to last week.
A British Library video curator in the aftermath of Peter Jackson upscaling and recolouring WW1 footage for They Shall Not Grow Old. Expect to see the same arguments trotted out again in the coming years (or months).
We hope you found this interesting. If you have feedback or comments email us or leave it below, and please forward this on to friends and colleagues who may also appreciate it.
See you soon!