The post-production pipeline, joint by joint

Post-production is usually described as a list of stages. The stages are rarely what makes it slow. The trouble lives in the joints between them, the moments when work changes software or changes hands and something quietly falls out. Here is the whole chain from card to archive, with every handoff named, what should cross it, and how you know it has broken.

A pipeline is its joints, not its stages

Ask any photographer to list the stages of post-production and you get the same answer: ingest, select, edit, export, deliver, archive. Nobody gets that wrong. Nobody's Sunday disappeared because they forgot editing existed.

The time goes somewhere less obvious. It goes into re-importing a folder because the selections did not travel. Into re-sorting inside Lightroom because the flags stopped at the previous app. Into hunting for a file whose name changed after you picked it. Into deciding the same thing twice in two different programs. Every one of those is a joint, not a stage, and joints are where systems leak.

So it is worth defining what a healthy joint looks like, because then you can inspect yours. A handoff is healthy when everything the next stage needs crosses it in one motion, and nothing has to be rebuilt on the far side. That is the whole standard. It says nothing about which software you use, which is the point: a pipeline is an agreement about what crosses each boundary, and the tools are interchangeable underneath it.

If you want the full business chain, from first inquiry through to archive and rebooking, that is our photography workflow guide. This piece narrows to post-production and stays there, and it deliberately looks at the gaps rather than the boxes.

Seven handoffs, card to archive

Here is the whole chain in one table. Read the right-hand column first. If any of those symptoms describe your last shoot, you have found your leak, and it will be more useful to fix than anything you could do inside a stage.

Seven joints, and in practice most photographers leak badly at exactly one or two of them. Finding which is worth more than a general resolution to be more organized.

Card to disk, disk to selection, pass to pass

Joint 1, card to disk. The contract here is simple and absolute: nothing gets judged until it exists in two places and you have confirmed the count. This is the only joint where being slow is correct. Verify, then proceed. Everything downstream assumes the files are safe, and if that assumption is wrong the rest of the pipeline is elaborate decoration on top of a risk. The sequence for the hour you get home, card reader to verified second copy, is in the after-the-shoot ingest checklist.

Joint 2, disk to selection tool. What should cross is the entire shoot, in capture order, as a single project, including the second shooter's cards merged in by time. Photographers who review card by card end up judging the ceremony without having seen it whole, which is a worse judgment made more slowly. This is also the last honest moment to rename anything. File names and folder structure are decided once, here, and are then frozen for the life of the job, which is the point of having a file naming system at all.

Joint 3, technical pass to story pass. The first pass is mechanical: focus, eyes, obvious misfires. The second is editorial: which version of this moment tells it best. What has to cross between them is not just a shorter list but the groupings, because the story pass is a series of comparisons and comparisons need the candidates side by side. Lose the grouping here and the second pass becomes guesswork against memory.

Selection to edit: the one that breaks most

If your pipeline has one bad joint, this is statistically where it is. It is the only handoff in the chain that usually crosses an application boundary, and application boundaries are where metadata goes to die.

Three failure modes account for nearly all of it.

The verdict never left the tool. You made a thousand decisions and they are stored somewhere the editor cannot see. For RAW files, ratings and color labels travel in the sidecar file that sits alongside the image, or through the selection tool's own export to your editor. If neither is happening, your decisions are trapped, and you will rebuild them by eye. Getting this right is the whole subject of exporting selections into Lightroom.

Identity changed in transit. A pick is a pointer to a filename. Rename, restructure or re-copy after selecting and the pointer dangles. This is why joint 2 insists on naming once, up front.

The decision was never finished. Frames arrive at the editor still marked maybe, so the selection quietly continues inside the develop module, at editing prices, with sunk cost pushing every marginal frame toward yes. That argument is the whole of why your picks set the edit clock, and it is the single most expensive habit in post-production.

One thing worth stating clearly, because plenty of workflows still assume otherwise: none of this requires anything to be thrown away. A clean handoff is a list of decisions crossing a boundary. The frames you did not pick stay exactly where they were, and the choice of what to do with them later is a separate question entirely. Whether you use flags, stars or colors to carry the verdict matters far less than picking one and never letting it carry a second meaning.

Edit to export, export to client, and the loop back

Joint 5, editor to export. The fix here is embarrassingly small and almost nobody does it: name your export presets after destinations, not settings. "Client gallery", "Instagram", "Album lab", "Blog". A preset called sRGB 2048 quality 80 forces you to re-derive the decision every time. A preset named after where the files are going is a decision you made once, correctly, when you were thinking clearly.

Joint 6, export to client. What crosses is a finished gallery and one message. The message matters more than photographers expect, because delivery is the moment the client re-evaluates the whole experience. A gallery arriving with no framing gets scrolled; a gallery arriving with a sentence about what to look for and what happens next gets shared. The mechanics of doing that well are in the client gallery delivery guide.

Joint 7, delivery to archive and back office. This is the joint everyone skips, and skipping it is why working drives fill up and invoices age. A job is not finished when the gallery goes live. It is finished when there is a cold copy stored away from the working drive, the final payment is settled, and you have asked for the review while the client still feels the thing they are reviewing. Batch this: one hour, once a fortnight, for every job that crossed the line since the last time. A cold copy that contains only the exports is not an archive, it is a souvenir: the originals and the edit settings have to go with it.

The two rules that turn steps into a system

You can build a working pipeline out of almost any combination of software, but only if it obeys two rules. Both are about direction and ownership, not features.

One direction of travel. Work moves forward, never back. Any step that sends work upstream is a leak with a friendly face: re-sorting after editing has begun, re-culling because you were not sure, going back to ingest to fix names, exporting twice because the first export went to the wrong place. When you catch yourself moving backwards, do not just fix the instance. Ask which earlier joint let it happen, because the answer is always an earlier joint.

One source of truth per fact. For every question the pipeline can be asked, exactly one place holds the answer. Where does "this frame is picked" live? Where does "this job is delivered" live? Where does "this client is paid" live? If a fact lives in two places, one of them is already wrong and you cannot tell which. This is the rule that turns a personal habit into something a second shooter or an editor can follow without asking you, which is the real test of whether you have a system or just a routine you happen to remember.

Everything else, tool choice included, is negotiable underneath those two.

Measure the joints, not the stages

Most photographers who try to speed up post-production start by timing their stages, and it rarely tells them anything actionable, because the stages feel like work and the joints feel like nothing. Time the joints instead. Four stopwatch numbers per shoot is enough.

1. Cards in hand to backup verified.
2. Backup verified to first selection decision.
3. Last selection decision to first correction applied.
4. Last correction to gallery live.

Do it for three shoots and one of those four numbers will be visibly, stupidly larger than you expected. That is your bottleneck, and it is almost never the number you would have guessed. Fixing the worst joint beats getting five percent faster at a stage, every time. There is more on diagnosing this in finding your post-production bottlenecks.

It is also worth noticing that joints 2 and 3 are the ones tied to a desk. They cannot begin until you are home, the drive is connected and the previews have built, which is why a Sunday shoot so often becomes a Tuesday problem. That constraint is a property of the software, not of the work. Kepla is being built to move those two joints onto an iPhone or iPad so the first pass can happen on the drive home, then continue on the Mac with the same shoot and every frame still present, because it picks and never takes a file anywhere. The apps are still in development. The booking page is live and free today while we build the rest.

FAQ