Archive-led documentaries have always been a popular staple that sell well around the world.
But the genre is under the same pressure as the rest of the industry, with margins squeezed and production costs up.
However, using AI in a strategic way can help unlock higher profits and save time by speeding up the production process, says Ben Field - the founder of Deep Fusion Films and one of the leading experts in TV about the new technology.
Although the company’s use of AI to re-create Michael Parkinson for the pioneering podcast series Virtually Parkinson was lauded, AI is now sometimes seen as a loaded word.
But Field stresses it can be used in beneficial ways if humans are in the loop. And it can actually improve not just the bottom line but also create new deals with archive libraries and distributors that give you a greater share of the money.
It could also benefit some production staff’s work-life balance and mental health, Field thinks.
For The Indie Hustle he breaks down how his company has re-shaped its way of working to use AI to its advantage and that of its staff. So read on to find out:
• The new economics
• How to pitch archive shows in reverse
• How you can create a faster post-production pipeline
• What responsible AI looks like
The new economics
One of the main reasons that there “isn’t enough TV being commissioned”, thinks Ben Field, is “because the cost of production is too high.”
Setting the scene, he points out baseline costs have been driven up by inflation - partly for laudable reasons, Field stresses, such as wage increases to counter the rise in the cost of living.
But the net result is it’s “now so extortionately expensive to make anything, that anything that is commissioned has to be so risk-free, that you end up regurgitating shows that have been around forever.”
Meanwhile, there is a “very good reason why YouTube is doing as well as it is”, because many creators aren’t charging for their own time - they’re just trying things out on their own dime and “audiences are responding to that” because that is where “edgier” new content is coming from.
The traditional TV funding model for archive also leaves producers with less to show for international success. Fifteen years ago global sales would have paid off but now says Field producers can often end up with “a profit margin of between 8, 9 or 10%, if you’re lucky”.
He says indies “can’t sustain” that level, so Deep Fusion began experimenting to see if it could use AI, “to create a system that can analyse and understand all forms of material…photograph, audio [and] video [so] we could then use it in an edit [and] essentially prompt our way through an edit.”
Field tested it by making a 45-minute documentary about Gerry Anderson, a subject he knows well having made the film Gerry Anderson: A Life Uncharted about the famous Thunderbirds producer.
Crucially, “human craft and creativity and skill and judgement were in the loop because I know the material”. He discovered he “was able to make a 45-minute edit in three days.”
That speed doesn’t apply to every genre. (“I’m not advocating this for investigative journalism” Field notes) but Deep Fusion began thinking how it could be applied to archive. Field discussed the possibilities with Getty Images’ Simon Wood, whom he already knew, on ways of using AI to improve the process - more on which later.
Field acknowledges “some people do sort of now, unfortunately, freeze when it comes to the word AI.”
But he explains: “The reason that we joined John Gore Studios was because we saw that there was the possibility to completely reimagine...traditional production. And that is not to remove any of the human control, authorship, craft and skill, but rather to empower people to be able to do things in a more efficient manner.”
“We are now capable of producing premium content at a price that is frankly more than affordable for distributors to sell globally...We’re mitigating risk and we’re introducing a new way of working to the marketplace.”
Takeaways: think where AI can speed up your systems; be realistic about what margins you’re left with; look at where audiences are finding new, riskier content
How to pitch archive shows in reverse
One of the ways Deep Fusion has been exploring making archive more profitable is by reverse engineering the process.
Usually with an archive doc, Field explains, a producer has an idea and “they’ll do a little bit of light research on it because there’s no money to do any full development on it”.
Then if a commissioner says yes, “you employ a researcher for three weeks to go away and find all your archive.”
His team decided to flip that system on its head: “We want to know what archives are out there in the world already and then we want to be able to pitch the show knowing the stuff’s there.”
To illustrate the issue, Field explains he searched Getty’s archive for footage of the D-Day rehearsal landings near where he lives. It came back with 38,000 results, most of which were irrelevant because the search simply picked up the word ‘rehearsal’.
“There’s no contextual understanding of what’s happening at all,” he says.
So Deep Fusion asked itself, “can we build software that understands archive?” and spoke with Getty about it.
Field explains his company then “built a tool that looks at all the archive from whichever library we’re looking at”. It now has “this archive search tool to be able to contextually understand what it is that we’re trying to make and then return only the results that help us tell that story.”
The company also wanted the tool to know “what we wanted to do” and “map out the story beats for us, which we could tweak and change based on the treatment that we put in” and then “attach the right archive” to those story beats. It can execute that process in as little as 15 to 30 minutes.
That changes what you pitch, says Field: “We now know that these shows are viable because we’ve got the archive and we’re doing that as part of the pitch.”
It has opened the door to a new kind of deal; Field says Getty “said they want to be more collaborative. They’re happy to take a back-end position and so forth, as long as there’s some investment up front.”
So Field took the idea to distributors with a proposal for a three-way co-investment: “We will invest some money, you invest some money and Getty will offset some money.”
In return, the producer, the distributor and Getty each own roughly a third of the content and “take a back-end share for the life of the deal” which works out at roughly (with some deal mechanics in) around 33p each from every pound.
Takeaways: identify available archive before you pitch; talk to archive libraries about taking a back-end position; approach distributors with co-investment models that give everyone a stake in long-term sales
A faster post-production pipeline
Many companies are trying to speed up their production and post-production pipeline using AI.
Deep Fusion has its own post system called PaperCut, which analyses material, creates metadata and then helps build the edit.
Field says using its AI tools means Deep Fusion’s budget lines “are exactly the same as they always have been”, and it uses researchers, APs, a PD, editors, composers and producers etc. The difference is that the AI tools mean “you don’t need to spend as long doing the grunt work on this. You can now just spend the time doing the polish and the online, which is the bit that everyone enjoys doing.”
That requires experienced people at the helm. “We still need people who know what they’re doing. If we handed our tool over to somebody who had never made a programme before, they wouldn’t know if the programme was right or wrong. We still need experts in their field.
“What you’re after in somebody’s expertise is judgement”, he explains, and their “understanding of what questions to ask in the first place”.
Field reckons AI can also be used to make programmes better. He left PaperCut running overnight on a project and it added 15 minutes to the edit. He cut half of it, but “found seven and a half minutes of good material that I had missed. It’s actually improving what we’re making.”
With the help of its AI tools, Deep Fusion believes that, although every deal is structured differently, in some instances profit margins could be trebled.
There are wellbeing benefits too. Because processes “can be left running overnight”, Field says, “we can stick to our 10-hour days…We’re not doing 18-hour days.” Also, because the system gives a written version of the edit first (hence its name PaperCut) it describes what’s in the footage so can also warn staff about distressing material before anyone has to watch it.
Takeaways: use AI for “grunt work” so your team can focus on craft and judgement; schedule workflows to run overnight to protect people’s working hours; use AI paper edits and content flags to protect staff from distressing footage
Using AI responsibly
Despite being AI-native, Deep Fusion is not using AI on screen in its forthcoming film Hunt: We Need to talk About James, about the late racing driver James Hunt.
Field explains why: “The James Hunt audience is going to be people who were probably in their 20s in 1975,” he says. “They are going to be the people that are probably least receptive to AI appearing in a documentary.”
His team also ruled out animated sequences as, “even if we use traditional techniques, somebody would think we’re using AI.”
The same thinking has been applied to a new AI tool that is available to programme-makers which cleans up scratches and blurriness on old footage. Field expects it could be used on Second World War material, but not on the Hunt film because “people want that tangible analogue feel because it is reminiscent; it’s nostalgia”.
“Just because you can do something it doesn’t mean you should. It’s all about choice,” he reckons.
Field also stresses that “even if we use an AI to do a first pass on any level of research, which naturally we do, we ask explicitly for record keeping and references and so on, so that a human can go back in and at the point that we know we’re going to use something in a programme, there’s a paper trail that says exactly where that came from.”
Deep Fusion is now so invested in AI, it has its own technology division and has appointed a head of software, Sam Hodgkinson. Plus it has a range of tools that it’s selling to other indies through a monthly tiered subscription.
Field says the four new tools they’ve recently released, which include Scripted PlannR (turns scripts into shoot plans) and MusicQ (identifies music and its cues) “come from somebody saying ‘I don’t want to do that’,” whether that’s music cue sheets or production paperwork.
Of course AI can now be used to vibe code your own tools, although not all indies will have the time to develop and test them, as Deep Fusion has done.
The technology also enables rivals to copy others’ tools but Field isn’t unduly concerned, as he points out big companies will already have their own, but also his priority is “making life easier for indies to survive”.
Takeaways: evaluate not just what AI can do but how your audience will feel about seeing it used on screen; be transparent about how you use AI; consider using affordable off-the-shelf specialist tools or creating your own to cut admin time
Next week's Indie Hustle will be published on Tuesday and cover Indielab's Growthlab conference.
Hiring? We feature paid job listings, to book a listing, email hello@theindiehustle.com
Sponsorship? If your organisation would like to find out about sponsorship and advertising opportunities, email hello@theindiehustle.com