Brief to published: the give-it-away creative workflow

By Dominique Davis · July 10, 2026

Here is our entire creative workflow, from a one-line brief to a published, multi-format piece, with the tools named and the steps in order. No held-back step, no "contact us for the rest." If you read to the end you will have the method we actually run. This post is the give-it-away version of how a piece of content goes from an idea to live.

What you'll learn

Why hiding the method is a weaker business than giving it away
The full pipeline, stage by stage, with the real tools at each step
Where the model does the work and where a person must
The checkpoints that catch a bad piece before it ships
What an end-to-end, given-away workflow makes possible

The old way: a black box and a long wait

Traditional creative production is a black box with a queue at the front. You hand in a brief, it disappears into a building, and weeks later something emerges that may or may not be what you meant. The process is opaque on purpose, because opacity is how the bill stays mysterious.

The cost is not just money. It is the feedback loop. By the time you see the work, the moment that prompted it has passed, the brief has drifted, and revising means re-entering the queue. Slow, hidden, and hard to steer.

The instinct, when you have a fast method, is to hide it too. We do the opposite, and the reason is practical, not noble. A method you give away proves you can run it, and the proof is what earns the work. People do not hire the studio that guards a secret. They hire the one whose method they have already watched work.

The new way: one brief, one spine, many outputs

Every piece starts as a one-line brief: the claim, the audience, and the place it lands. That single line becomes the spine that every later stage reads from, so the piece stays coherent as it splits into formats.

Here is the full pipeline.

1. Brief to outline

We expand the one-liner into a tight outline with Claude, then cut it by hand. The model is fast at structure and bad at knowing what matters to us, so it drafts and we decide. The outline is the contract. If it is wrong here, everything downstream is wrong cheaply, which is the right place to be wrong.

2. Outline to script

The outline becomes a script in our voice. We feed the model our voice rules and prior pieces, it drafts, and a person rewrites the open and the close, because those are the two places a piece lives or dies and they are exactly where a model is weakest.

3. Script to voice

The script goes to **ElevenLabs** for a voiceover in our locked voice. One narration, generated in minutes, that every video format will share. Consistency for free, because it all comes from one source.

4. Voice and visuals to video

We render the visuals in **Remotion**, our motion system, timed to the narration. Stills and any generated art come through our image pipeline. **FFmpeg** muxes the audio to the picture and produces the master cut.

5. Master to formats

From the master we derive the formats: vertical cuts, captioned social edits, a written version for the blog, an email. Each reads from the same spine, so the claim stays identical across every surface even as the framing changes per channel.

6. Formats to published

The pieces go out through the channels they were cut for. The record of what shipped, where, and when goes back into **Airtable**, so the next piece starts with a clear picture of what already exists.

Where the model works and where a person must

The honest map of this workflow is a map of a division of labor.

**Model does:** structural drafting, the first pass of a script, the voiceover, the mechanical rendering and reformatting. The repetitive, fast, low-judgment work.
**Person does:** the brief, the outline decision, the open and close, and every quality checkpoint. The high-judgment work that the whole piece hangs on.

We never hand the judgment to the model, and we never make a person do the mechanical reformatting by hand. Putting either in the wrong column is how the workflow rots, either into generic output or into a person buried in busywork.

The clearest sign a team has the columns wrong is what their output sounds like. When the model is doing the judgment, every piece reads the same, smooth and forgettable, because the model defaults to the average of everything it has seen. When a person is doing the mechanical reformatting, the team is slow and resentful and the volume never grows. The right split produces the opposite: pieces that sound like a specific point of view, shipped at a pace a hand-formatting team could never match. We watch for both failure smells on our own output, because drifting into either is easy and only the work tells you it happened.

A worked example: this week's social piece, traced

To make the pipeline concrete, here is how one real piece moved through it this week.

The brief was one line: "show that one finished story becomes nine native social edits, with the tools named." That line went into the outline pass with Claude, which returned a structure of six sections. We cut two, merged two, and rewrote the order by hand, because the model put the worked example last and we knew it needed to come earlier.

The outline became a script. The model drafted the middle in our voice rules; a person rewrote the open, which originally started with a definition, into a flat statement of the failure it fixes. The close, which the model ended on a summary, got rewritten to circle back to the opening line. Open and close, human, every time.

We did not need a voiceover for the written version, so we skipped the ElevenLabs and Remotion stages for this particular piece, which is the point of a pipeline built from one spine: you run the stages a given format needs and skip the ones it does not. The written piece went through the publish checklist and out. The same spine is what the video edits will read from when those formats render.

The whole thing, brief to published draft, was an afternoon. Most of the afternoon was the two rewrites that mattered, the open and the close, and the publish check. The mechanical stages took minutes and ran while we did the thinking. That ratio, most of the time on judgment and little on mechanics, is the whole design.

The checkpoints that catch a bad piece

Speed without checkpoints just ships bad work faster. We gate at three points.

**After the outline.** Is the claim true and worth saying. Kill it here if not, before any production cost.
**After the script.** Does it sound like us, and does the open earn the rest. This is a human read, out loud, every time.
**Before publish.** A plain checklist: no banned words, no fabricated numbers, the links work, each format fits its channel. We run this on our own pieces with no exceptions, including this one.

We built this pipeline running it on Temerarii Media's own daily content, which is why we trust it: the failures showed up on our own work first, and we fixed the workflow before anyone else's content ran through it.

What this makes possible

An end-to-end workflow you can hand to someone, in full, changes the relationship. The work goes from idea to live in a day instead of weeks, and you can steer at every checkpoint instead of waiting for a reveal. The cost moves from the production, which is now mostly scripted, to the thinking, which is where it belongs.

And giving it away does the quiet selling. A reader who runs this method themselves and hits the part where judgment matters, the open, the close, the taste, often decides that is exactly the part worth bringing to a studio. The method is free. Running it well, consistently, on your real work, is the thing.

So there it is, the whole pipeline, no withheld step. Brief, outline, script, voice, video, formats, publish, with a human at every gate that matters. Run it yourself, or watch it run in public.

The experiments at [temerarii.xyz](https://temerarii.xyz) are this workflow working where you can see it, and the rest of the [blog](/blog/) takes each stage apart in more depth.

See the live system →

Written by Dominique Davis, Founder, Temerarii Media.