Module 02 · Scripting for AI Video
Adapting Scripts for Roman-Urdu and English Audiences
Open lesson + course map
On this lesson
Course outline
Module 1 · The AI Video Stack in 2026
Module 2 · Scripting for AI Video
Module 3 · AI Voice and Narration
Module 4 · Generative Video With Veo
Module 5 · AI Avatars and Presenter-Style Video
Module 6 · Editing and Assembly in CapCut
Module 7 · Video Service Packaging
An English script and a Roman-Urdu script can carry the same promise without using the same words. Good adaptation preserves the facts, sequence, and call to action while changing idiom, sentence length, examples, and spoken rhythm for the people who will hear it.
By the end, you will have one matched script pair, a record of the language decisions you made, and a review method that catches changed meaning and awkward text-to-speech pronunciation before production.
// concept
Choose the Language for the Audience, Not the Creator
Roman Urdu can suit a conversational Pakistan-focused video; English may fit international vocabulary or a multi-country audience. Test rather than assuming either will gain more reach.
Use the video destination and viewer task to choose a starting version:
| Niche and intended viewer | Sensible first version | Why | What to test |
|---|---|---|---|
| Tutorial for Pakistani first-time freelancers | Roman Urdu plus familiar work terms | Conversational explanation; client vocabulary stays intact | Completion and clarification questions |
| Local WhatsApp retail explainer | Roman Urdu or viewers' preferred language | Spoken familiarity matters | Matched uploads or viewer review |
| Global software walkthrough | English | Interface labels remain consistent | Comprehension with/without localized captions |
| Editing portfolio for foreign clients | English, optional localized captions | International buyer context | Qualified enquiries and retention |
For a fair test, keep visuals, duration, offer, and publishing window similar. YouTube may support extra audio tracks and analytics by audio language; verify current eligibility and controls.
// concept
Build a Meaning Contract Before You Adapt
Mark what must remain identical in a four-row meaning contract:
| Element | Source decision | Adaptation rule |
|---|---|---|
| Core claim | Review scope, deadline, and revisions before accepting a client brief | Keep all three checks in both versions |
| Evidence | No outside claims | Add no authority language or outcomes |
| Order | Problem, three checks, action | Preserve the sequence so visuals can be reused |
| Call to action | Save checklist before replying | Preserve the action |
Then create a term card. Roman Urdu has no single spelling convention, so use one consistent house style per video.
| Source term | Chosen Roman-Urdu form | Decision |
|---|---|---|
| scope | scope | Explain once as kaam ki had for beginners |
| revision | revision | Use revisions as the familiar plural |
| agree | haan kehna | More natural here than the literal raazi hona |
Use this prompt only after you have written the meaning contract:
You are a bilingual script adapter, not a word-for-word translator.
Audience: Pakistani freelancers using conversational Roman Urdu and common
English work terms. Delivery: warm, short spoken sentences, 45-60 seconds.
Preserve exactly:
1. the three checks: scope, deadline, revisions;
2. the order: problem -> checks -> action;
3. the CTA: save the checklist before replying;
4. every qualification in the source. Add no facts or results.
Term card: scope -> scope (explain once as kaam ki had); deadline -> deadline;
revision -> revision; agree -> haan kehna.
Adapt idiom, pacing, and sentence boundaries for natural speech. Do not force
slang. Return: (A) Roman-Urdu script, (B) line-by-line meaning back-check,
(C) terms whose spelling or pronunciation needs a human test.
<source_script>
[paste the approved English source]
</source_script>The back-check is a literal account of what each adapted line now means. Compare it with the approved source to catch missing or stronger claims.
// concept
Review With a Back-Check, Read-Aloud, and TTS Audition
Run three different reviews; they catch different defects.
- Meaning back-check: Translate the adaptation's meaning into plain English. Compare claim, qualification, order, and CTA; restore omissions and remove additions.
- Human read-aloud: Have a fluent target viewer read without rehearsal. Mark stumbles, breathless sentences, and phrases they replace naturally.
- TTS audition: Generate only 6–10 difficult lines first. Test names, numbers, and code-switching on a phone speaker.
Roman-Urdu TTS may misread a sensible spelling. Record tested substitutions such as mujhe -> mu-jhay; fixes are voice-dependent. ElevenLabs documents dictionaries and aliases, but model support differs. Verify it, respell one sentence at a time, or use a consenting human speaker.
Save tts-test-a.mp3, tts-test-b.mp3, and note the single changed spelling. Choose by intelligibility, not dramatic delivery.
// worked_example
Worked Example
This is a fictional sample for a Pakistani freelance-education Short.
Approved English source
Before you accept a client brief, check three things. First, define the scope:
what exactly will you deliver? Second, confirm the deadline and the client's
review time. Third, write down how many revisions are included. A vague brief
can hide extra work. Save this three-point checklist before you reply.A literal first draft produced: Client brief ko qubool karne se pehle teen cheezain jaanchein... Mubham brief izafi kaam ko chhupa sakta hai. Its formal wording and English sentence shapes sound dense in a conversational video.
The adapter retained common work terms and split the middle into breathing units:
| English meaning unit | Adapted Roman Urdu | Adaptation note |
|---|---|---|
| Before you accept a client brief, check three things. | Client brief ko haan kehne se pehle teen cheezein check karo. | Natural speech replaces literal qubool karne. |
| First, define the scope: what exactly will you deliver? | Pehle scope clear karo: aap ne exactly deliver kya karna hai? | Keeps the client term; changes spoken order. |
| Confirm the deadline and the client's review time. | Phir deadline ke saath client ka review time bhi confirm karo. | Preserves both timing checks. |
| Write down how many revisions are included. | Aur likh lo ke kitni revisions package mein included hain. | Adds context, not a new promise. |
| A vague brief can hide extra work. | Brief clear na ho, to baad mein woh kaam bhi aa sakta hai jo likha hi nahin tha. | Re-expresses hide naturally. |
| Save this three-point checklist before you reply. | Reply karne se pehle yeh teen-point checklist save kar lo. | Same action and order. |
Back-check: clarify deliverable, deadline, review time, and revisions; an unclear brief may lead to unwritten work; save the checklist before replying. No meaning or result changed.
For TTS, audition cheezein, revisions, and included. If one fails, test a respelling or rewrite that phrase; do not assume the fix transfers to another voice.
// failure_cases
Failure Cases to Diagnose
6 cases to diagnose
Literal but unnatural register
valid words such as
mubhammay sound too formal. Let a target viewer replace them aloud.Meaning drift
the Roman-Urdu version changes “may” into “will” or adds a result such as “you will avoid bad clients.” Back-check modal words and delete the new promise.
Uncontrolled code-switching
every noun becomes English. Keep real work terms; replace code-switching that only signals style.
Inconsistent Roman spelling
captions show
cheez,cheezeandcheezainin one minute. Choose a house spelling in the term card and apply it to script and captions.TTS-first editing
the writer distorts natural Roman Urdu merely to satisfy one synthetic voice. Test another spelling or voice, and prefer a human recording when intelligibility remains poor.
Invalid language test
uploads have different hooks, lengths, or visuals. Match non-language variables first.
// pakistan_angle
Pakistan Angle
Pakistan is multilingual; Roman Urdu is not automatically the best local choice. Viewers may be more comfortable with Urdu script, Punjabi, Pashto, Sindhi, Saraiki, Balochi, Hindko, English, or a particular mix. Do not turn city, province, or accent into a shortcut for presumed language. Ask the client which language their customers actually use, then test the script with two or three intended viewers before recording the full track.
Roman-Urdu spellings also vary across Pakistani WhatsApp messages and captions. For a mobile-first video, pick readable spellings, burn in carefully proofed captions, and check them on a modest Android screen. During load-shedding or limited data, exchange the script and a compressed voice test first; upload the full-resolution edit only after language approval so a pronunciation correction does not trigger another large render and transfer.
// hands_on
Hands-On Exercise
8 steps
Produce one English/Roman-Urdu script pair from a 45–60-second script you own or are authorized to adapt.
Name the audience, market, platform, and viewer action.
Build a four-row meaning contract covering claims, evidence, order, and CTA.
Create a term card with at least eight decisions: keep, explain, replace, or respell.
Use the adaptation prompt, then edit the Roman-Urdu result by hand.
Back-check each line and correct changed or missing meaning.
Read both scripts aloud and revise any line that cannot be spoken comfortably in one breath.
Audition 6–10 difficult Roman-Urdu lines with your intended TTS voice, or with a consenting human reader. Record pronunciation fixes.
Save both scripts and notes explaining three non-literal choices. You are done when both scripts communicate the same claims and CTA, each sounds natural aloud for its audience, and another reviewer can trace every adaptation decision.
// completion_rubric
Completion Rubric
6 checks — tick as you verify
// sources
Sources
4 official sources — check every claim yourself
// check_yourself
Check yourself
4 questions · answers and options are taken word-for-word from this course
1 / 4 · diagnose
Your work shows this failure mode: “Uncontrolled code-switching.” The lesson describes it like this: “Every noun becomes English.” What does the lesson tell you to do about it?