Eric (Yichen) Zhang

A vertical drama platform releasing more than sixty titles a month. I joined as an assistant, was promoted to Costume Designer after ten, and ran the department for the rest of my time there. Four titles were mine end to end, fifty to eighty episodes each.

At that speed you cannot design by taste alone. You need a chain that survives being handed to other people. This chapter is that chain, one link per section: read the script into states, set the character, prove it on the body, hold it on set, and then look at what the audience actually said.

Titles here are working titles, replaced with placeholders. Released titles differ.

“Soft but sharp — clean fit, neutral colours.”

She is not poor, she is unplaced. Clothes that fit properly in colours that decline to announce anything, so the room can read her as staff or as family depending on what it wants.

“Muted, soft colours, washed off — showing the fragile side.”

Saturation is the tell. Everything is one wash down from where it should be, so that when the palette finally comes up it reads as a person arriving rather than a costume change.

Three passes per character. First the silhouette: what shape the audience should recognise at thumbnail size, because on a phone a viewer decides in half a second. Then the fit, which is where status lives: the same jacket reads as inherited or borrowed depending on where the shoulder sits. Then colour, held back deliberately so the arc has somewhere to go.

Notes were written to be executable by someone who had not read the script: a stylist pulling on a Saturday, or an outside vendor with no context. A reference that only works if I am in the room is not a reference.

Reference boards contain third-party imagery collected for internal use and are not reproduced.

“The reader is a stranger, in a hurry, working in the other language, and will not ask a follow-up question.”

Sixty titles a month, a department of two, and most titles shot by outside partners. Past a handful of titles the usual mechanism for holding a look together, a person on the floor making the call, stops scaling. Consistency has to be carried by a document instead of by supervision.

So it is not written to be read front to back. It is written to be entered: a reader arrives with a specific problem and should reach a resolved answer in one or two steps. Four separate indexes run over the same material.

It replaces judgement with lookup. A partner does not have to understand our taste, absorb a mood board, or guess what we would have wanted. They locate the character type, the gender and the occasion (or the scene they are shooting) and land on a cell already resolved into finished looks. The interpretation was done once, centrally, and then spent hundreds of times.

It is also why the archetype library is images rather than prose. A written description of a silhouette survives translation badly and invites negotiation. A board of assembled looks does not.

The rules are set down in English and in Chinese, not as a translation appended afterwards, but as two parallel statements of the same instruction. Execution was split across two languages and two time zones, and a rule that exists in only one of them is a rule half the partners are guessing at.

Background actors and service staff get the same treatment as leads. There is a section on what survives being wet and shot underwater, and guidance on preparing performers for intimate scenes. None of that is styling; all of it is what stops a day from falling apart.

Apparel for the company itself: one tee block, differentiated by role. Cast, crew, director, producer, and a bilingual edition for international shoots. Each in black and white, so a set can be read at a glance and a photograph of that set stays on brand. Laid out below in the order the range was built: role first, black before white.

The vest is the piece I would defend hardest. Directors work a whole day out of their pockets, so the pocket plan came first and the garment was drawn around it: chest pockets for what gets reached for standing, lower pockets sized to actual objects, role designation across the back where it reads from behind at distance. Three variations were drawn against the same block, differing in how the pocket bank is split.

At the end of the run I put together a report arguing the department's case in the platform's own terms. In short-form drama the commercial risk is looking cheap, so the question is not whether the costume is good. It is whether the audience says the show looks expensive.

Across the sampled posts, wardrobe showed up unprompted in the comments more than a dozen times, and the recurring words were the ones that matter to a platform selling episodes.

Costume stopped being background and became something viewers talked about, which is the only argument for a wardrobe budget that a platform actually has to accept.

• “This looks premium”
• “This looks expensive”
• “Obsessed with the styling”
• “The outfits are everything”
• “Whoever did the costumes is killing it”
• “The uniform suits you”

Public comments, quoted without attribution. Account names and images withheld.