QWE AI Academy Here’s the math most one minute episode guides skip: natural speech sits around 140-160 words per minute. For a full 60-second episode that means roughly 120-180 total words – dialogue plus short action cues. Cross ~200 and TTS rushes, or you chop emotion mid-line. (Timing ranges lined up with Biteable’s video word-count guidance and other 2025-2026 script-timing writeups.)
Phone apps already run on that constraint. MOCHI.TV ships vertical anime-style series as one-minute hits – full-season stakes, daily drops, cliff every exit. Same box, different job: you + ChatGPT + a free generator, no crew.
Think of the clock as a match. Light it, burn bright, leave them needing the next spark. Long setup is how retention dies.
Word budget before pretty frames
One minute episodes aren’t short movies. One dramatic turn: cold open, one escalation, one reveal that reframes the open, hard cliff. Vertical 9:16. Two or three characters. Every extra syllable steals face-time on screen.
Simple language isn’t style preference – it raises achievable WPM without panic pacing. Action in brackets is silent time. Budget it or the read-aloud blows past 55s.
Script → generate (with the cap enforced)
Beat sheet first. Frames second.
- One-sentence premise with imbalance already inside it. Example: “A night-shift barista discovers every customer she serves has already died – except the one who tips in blank envelopes.” Not billionaire-CEO revenge. Imbalance does the hook work.
- Force a timed script. ChatGPT or Claude. Paste and swap brackets:
Write a vertical 9:16 one-minute episode script. Hard max 160 words total including brief action.
Exact timing:
0-8s HOOK: cold open, one shocking image or line, zero backstory.
8-25s ESCALATE: one obstacle + clear emotional cost; dialogue under 40 words.
25-45s REVEAL: single twist that re-frames the first 25 seconds.
45-60s CLIFF: unanswered physical or emotional threat that requires episode 2.
Format: NAME in CAPS, short (action), then dialogue. 2-3 characters only.
Genre: [urban mystery / found-family sci-fi]. Premise: [your sentence].
Reuse these exact visual tags every time: [Character A: age, 3 anchors - hair, coat color, one scar or prop].Phone timer. Read aloud. Over 55 seconds? Cut adjectives, then whole lines. Don’t generate until the read fits.
- Generator that hard-caps at 60s.Crreo Free (as of latest public help-center/brand-kit copy: $0, no card) maxes at 1 minute, watermarks exports, text-based characters only, tight monthly minute pool – community and docs often cite roughly ~5 video minutes total. Paste the timed script, 9:16, identical character strings, generate. Regen only weak clips; regens spend the same pool. Grow+ is where photo characters show up; Start is the lower paid rung (from roughly $14/mo or ~$9 annual per pricing pages – verify current numbers before you pay).
- Light assemble. Music under dialogue, not on top – phone speakers lose the fight. Export vertical. End card teasing episode 2.
Where free tiers actually break
Draft 250 words and the voiceover isn’t the villain. The budget was. Time the read before you touch a generator.
The catch is burn rate. Crreo’s free minute pool (see their help center) treats regenerations as real spend – watermarked tests still count. MOCHI.TV creation runs on gems; App Store reviews keep saying starter gems often won’t finish one full episode without a multi-day wait or a top-up (setting/character costs plus per-scene gem hits add up fast). Pippit Short Drama Agent (Seedance-powered) hands out daily/weekly free credits; agent runs can chew a large block in one pass – fine for a test, rough as your only pipeline.
Continuity trap: Free/text character modes have no photo lock and no cross-episode memory. Save one short three-anchor string (hair, coat, scar/prop) and paste it unchanged every generation. Change one adjective and faces/wardrobe morph by episode 3 – viewers clock it instantly.
Options at a glance
| Tool | Best for | Free reality (as of latest public info – recheck before shipping) | Catch |
|---|---|---|---|
| MOCHI.TV | Watching + in-app anime-style create | Free to start watching; gems for create/enable | Gems burn per scene; free progress often stalls |
| Crreo Free | One-shot 60s script-to-video tests | $0, 1-min hard cap, watermark, text characters | Monthly minute pool; photo lock needs Grow+ |
| Pippit Short Drama Agent | Script upload → multi-scene drama | Daily/weekly free credits | Agent mode can swallow big credit chunks |
Tools only render beats you already forced into the word cap. Season bibles and reference-image sheets matter after you leave free tiers – not before.
Honest question worth sitting with: if episode 2 already needs a paid gem pack or a minute top-up, is the premise strong enough that you’d pay to finish it yourself? If not, fix the hook before you fund the pipeline.
FAQ
How many words should a one minute episode script have?
120-180 total, action cues included. Matches ~140-160 wpm so the VO stays human.
Can I stay on free tools for a whole series?
Episode 1 watermarked? Sure. Clean multi-ep with stable faces and no watermark? Paid shows up fast – Crreo Start/Grow range or MOCHI gem packs. Concrete path: three locked text characters, five free 1-minute tests, then the monthly minute wall mid-regen on episode 3. That’s upgrade-or-switch, not “try harder on free.”
Do I need the same characters looking identical every episode?
For binge retention, yes. Face morph is obvious on a phone loop. The misconception: “the model will remember.” On free/text tiers it won’t. You own continuity – identical three-anchor paste, every scene, every episode. Photo reference on higher plans is easier; it still isn’t free. Skip the lock and the cast changes weekly. Cliffs won’t save that.
Open the beat prompt. Your premise. Timer to 55s. Free 1-minute generator. Episode 1 can ship today – if the read-aloud already fits.