A ninety-second episode has to seduce, shock and cliff-hang before the thumb moves. Most of that is measurable — the word budget of each block, how long before the first beat lands, how many locations you are paying for. Paste an episode below and find out in three seconds. The measuring is free and anonymous. The writing is not.
New to the format? The Vertical Drama booklet is free to read — fourteen chapters, no account. PDF and EPUB on request at ludovic@streaming-radar.com.
Numbers from an analysis of more than a hundred English and Chinese series, plus the paywall conventions the market actually applies. Nothing here is an opinion about your story.
The viewer decides inside fifteen seconds. Open mid-crisis: the slap has already landed. Studios put the most cinematic beat of the season here — it doubles as the trailer.
A jolt every forty seconds at most: a reveal, a reversal, a burst of dread. No connective tissue — compress the non-drama, then punch it.
Fewer words. The last image is held, not spoken — long enough for the actor to become the episode thumbnail.
150 words in 15 seconds is 10 words a second. A hundred words across the last thirty is 3.3. Averaging the two betrays both ends: an under-written hook and a chatty cliffhanger. So the budget carries three separate rates, and the hook and cliffhanger windows are absolute — they do not stretch when the episode runs to five minutes. The escalation absorbs the rest.
Four can be measured on the text: a lead who swallows the dialogue while someone else exists only to hand over a prop, machine-translation residue, cliché density, and backstory dumped into voice-over because there was no room for a scene. The fifth — twist addiction — cannot. A gratuitous reversal and a planted one are identical to a counter; you have to read the set-up. The audit says so rather than pretending otherwise.
Paste a script. Nothing is stored, nothing is sent to a model, no account needed.
Standard screenplay form works best — sluglines in INT./EXT., character names on their own
line, and timecode markers like [0:38] if you have them, which is what lets the pacing be measured.
| Measured | Value |
|---|
Measuring is free because it costs nothing to serve. Writing runs a model, and it draws on the
Vertical Invasion 2026 silo — so it sits behind the MCP server, with an API key and the
vertical-invasion-2026 entitlement.
| Output | What you get |
|---|---|
beat_sheet |
The season skeleton, episode by episode: Beat / Turn / Fall, the paywall spine, and the list of episodes to dialogue next. This is what platforms read first — before any dialogued page. |
episode_script |
A shootable episode written against the template above, with a timecode marker on every beat. It comes back already measured: the tool audits its own page before handing it over. |
bible |
The full series bible — logline, world, characters, three-act arc, episode breakdown, production and market notes. |
script_audit |
The same measurement as above, plus a script-doctor pass on what counting cannot see — which reversals have no set-up, whether each visual tell reads on camera. |
Free, no account. Pick what you want and get it in about twenty seconds. An episode script comes back already measured against the template above.
The same four outputs run through the Streaming Lens MCP server if you would rather drive them from your own agent — along with the market this writes for: audience by country, production costs, the platform catalogue, weekly signals. The template is free; the numbers are where the work is.