I was reviewing a passage in a ten-minute film when I realised the images had lost their connection to the voiceover. The clips felt too random for what the script was saying. I knew the feeling I’d had when I wrote it, and I wasn’t seeing that feeling on screen.
I stopped and talked it through. I described the mismatch and asked for the passage to be made again in a style closer to what I had imagined. New video clips were generated and placed in the edit. I watched the passage again. It had more of the speed and emotional feel I wanted, and the imagery belonged more closely to the voiceover.
The significant part, for me, was what I could hold together while doing it. I was looking at the film as it existed, listening to the music and voiceover, and remembering the feeling that had made me write the script. I could speak about the distance between those things while I still felt it. The edit could then change, and I could judge the result.
That is a different creative process from writing a brief and waiting to see what comes back.
A thought in motion
My creative thinking doesn’t usually arrive as a complete instruction. I might start by saying a cut needs to be faster. As I talk, I realise the timing may be right but the shot is wrong. Then I listen again and recognise that the real problem is larger: the sequence has drifted from the emotion of the voiceover.
Speaking lets me follow that discovery as it happens. I can offer an idea, correct it and become more precise without leaving the conversation. I am more descriptive when I talk than when I type. I find myself talking about how a shot should move, what the transition should feel like, where the energy should rise, and how an image sits against the music.
Typing has its own creative value. It makes me stop and examine a thought. But it also asks me to turn a reaction into organised language before I can give it to a tool. When I’m watching an edit, that pause can break my connection with what I’m seeing and hearing. By speaking, I can stay with the film while I work out what I mean.
Over the past month I’ve almost stopped typing in these conversations. I use Wispr Flow to speak into Claude and ChatGPT. A recent Flow screen showed 38,241 dictated words. An earlier summary recorded use across 25 apps, at an average of 168 words per minute. Those numbers say nothing about the quality of my ideas. They do show how thoroughly speech has entered my working day. My family sometimes hears me talking and wonders who I’m talking to.
I’m often thinking aloud to the work.
Keeping the original feeling alive
There’s a particular problem in making a film that I think other creatives will recognise. You begin with a feeling or an intention. You write a script, find or make images, build an edit, add music, revise. At every stage the work becomes more concrete. It can also travel further from the thing you first saw in your head.
That distance is difficult to describe. An image can be polished and still be wrong for the passage. A cut can be technically smooth and still lose the emotion. In the film I was reviewing, the issue was the relationship between the clips and the voiceover. The images weren’t carrying the thought that had produced the script.
When I spoke to the system, I could return to that thought. I didn’t need to identify every replacement frame in advance. I could describe why the passage felt disconnected and what I wanted it to become, then review the newly generated clips in place. I could move between a specific decision about one shot and a judgement about the whole sequence.
That movement between scales matters. Editing software makes individual shots visible and controllable. Creative direction also involves sensing the shape of the whole piece: its pace, its emotional rise and fall, and whether an audience can feel a connection from one moment to the next. I want to keep both levels in view.
The faster this cycle becomes, the longer I can stay inside that larger judgement. A feeling doesn’t have to be filed away as a note for later. I can test it while the voiceover is still in my ears.
When direction and production become one activity
I’ve been reviewing individual clips on screen while directing changes to the edit through my voice. I’ve talked through movement, cuts, replacements and pace. I haven’t needed to reach for a keyboard shortcut or operate an editing control to drive those changes.
That is a remarkable technical achievement. But its effect reaches beyond convenience. The familiar separation between having the idea, explaining the idea, making the thing and reviewing the thing begins to narrow. I’m moving among those activities continuously.
At moments it feels as though my process has accelerated tenfold. That is how it feels, not a measured claim. What has changed is the time between sensing that something is wrong and seeing a possible answer. I can stay close to the production while remaining fully engaged with the creative question.
This doesn’t make every quick answer a good answer. It makes more answers available for judgement. I still have to decide whether the new imagery belongs, whether the pace serves the voiceover, and whether the sequence expresses the feeling I was trying to reach. Those decisions become more frequent as the tools become faster.
I think this may alter what creative control means. Knowing how to operate a timeline is a valuable skill. So is knowing that a passage has lost its emotional thread and being able to bring it back. As systems become more responsive to natural direction, those abilities can meet more directly in the work.
What is happening in the mind?
I’m wary of claiming that speech itself is inherently more creative. A study comparing spoken and written idea generation found that participants wrote down more ideas than they spoke in its particular task, while the originality of their ideas did not differ. Writing can help us hold, compare and refine thoughts. I recognise that benefit in my own work.
My experience involves a further element: the film responds. I speak, a change appears on screen, and I react to it. The evolving work holds part of the conversation for me. I don’t have to keep every alternative in my head. I can see and hear one, judge it, and use that reaction to find the next direction.
That makes me wonder whether the important psychological change is the continuity of the feedback loop. I can notice, express, test and judge without repeatedly changing my attention from the work to the mechanics of making it. The thought doesn’t have to arrive complete. It can develop through contact with the result.
There is a risk in this, too. I think aloud. I pause because I’m searching for something. I may say one thing and correct myself a moment later. A transcript can capture the words while missing the significance of the pause. An LLM has to distinguish a tentative thought from an instruction I want carried out. So far, it has handled that better than I expected, but creative systems need to get much better at making room for uncertainty.
I’ve also tested recording my screen, voice and movements while directing work. That may give a system more context than a transcript alone. Still, I don’t need to believe that a machine feels my emotion. What matters is whether I can express enough of my intention to test it, then retain the authority to say whether the result is right.
Research into AI-supported video creation has explored how creators can remain oriented, inspect generated material and move between parts of a project. That question feels increasingly practical to me. A creative system needs to support the full process of exploration and revision, including the moments when the creator changes their mind.
I’ve been speaking directly with people at frontier AI labs, including Tavus, about the future of interacting with machines. Tavus builds real-time AI PALs that can see, hear and respond. Those conversations bring me back to a question from my own work: can a system stay with a person’s unfinished thought, rather than rushing to treat every utterance as a command?
The creative role ahead
I can imagine this changing how other creatives work. Someone may be able to direct a rough film while watching it, reshape an image while describing what feels wrong, or move between writing, sound and picture in one conversation. More people may be able to test an idea before every technical part of the process is in place.
For experienced creatives, I suspect the change will be just as profound. Production has often required a handoff: to another specialist, another application, or another stage of work. Those handoffs can bring valuable expertise. They can also interrupt the connection between an intention and the person who first felt it. A more responsive process could let several people contribute their judgement closer to the evolving work, rather than requiring each thought to become a separate brief.
The responsibility for judgement grows with that freedom. Faster generation can fill an edit with plausible material. It takes a person to recognise when the clips have become random in relation to the script. The skill is in hearing that disconnection, finding language for it, and continuing until the film carries the feeling it needs.
There must also be room to stop. An immediate next version is tempting, but speed can make a piece restless. Sometimes the best creative decision is to watch without changing anything, to write down why a choice matters, or to return tomorrow. I want tools that help me preserve versions and remember the thinking behind them, as well as tools that respond quickly.
I’m still discovering what speaking does to my work. I know that it lets more of my thinking come out before I’ve tidied it for somebody else. In the film, it let me reconnect new images with an emotion I had carried since writing the script. I could listen, speak, see a change and keep shaping the whole picture.
The machine doesn’t need to feel what I feel. I need to be able to work with that feeling while it is still alive.
Colin Melville is an Executive Creative Director. These views are his own.
More from The Shift ↗