user_raj87:
Seeing too many “AI film” videos lately. Some look crazy good, some look… weird.
Can someone explain proper workflow? Not theory, actual steps.
indie_dir_arun:
I’ve been testing this last few months. It’s not magic, but it works if you do it properly.
I’ll write what I actually do (short films, 30–90 sec).
Step 1 – story first (dont skip this)
Most people jump into tools directly, thats why output looks random.
I usually draft story using
ChatGPT
But don’t copy paste. You need to edit manually. AI writing still feels generic otherwise.
Example:
write a 1 min story
break into 5–8 scenes
Keep it short. Long films still not working properly.
scriptwriter_meena:
+1 on this.
AI scripts feel “empty” if you dont tweak. Dialogues especially.
indie_dir_arun:
Step 2 – generate scenes (core part)
Tools I tested:
Runway ML
Pika Labs
Sora
You type prompt → get video.
But reality:
first output mostly unusable
need multiple attempts
vfx_nikhil:
Important point: generate short clips only (3–6 sec).
Trying to generate long scenes = more errors.
indie_dir_arun:
Step 3 – character consistency (big headache)
AI doesnt remember faces properly.
What helps a bit:
describe same character in every prompt
keep wording same
don’t change too many details
Example prompt style:
“35 year old man, short beard, black jacket, serious expression, cinematic lighting”
Repeat this every time.
Still not perfect, but usable.
film_student_arjun:
So no way to fully lock character?
vfx_nikhil:
Not fully yet. Some tools improving, but still inconsistent.
indie_dir_arun:
Step 4 – voice
Options:
record yourself (best)
AI voice
I tested
Synthesia
Decent for simple videos, but long emotional dialogue still sounds off.
sound_designer_rahul:
Voice alone is not enough.
Add:
background sound
ambience
subtle effects
Otherwise video feels fake immediately.
indie_dir_arun:
Step 5 – editing (this is where it becomes watchable)
Use something like
Adobe Premiere Pro
Work you need to do:
cut bad parts
fix pacing
match audio
basic color correction
seo_guy_rahul:
From content side, short videos perform better:
30 sec stories
mystery / sci fi / emotional
Easier to hold attention.
indie_dir_arun:
Actual workflow summary (what works for me)
write short story
split into scenes
generate multiple clips per scene
pick best ones
add voice + sound
edit tightly
Time taken: around 3–5 hrs for one decent video
freelancer_meena:
Clients already asking for this type of videos.
Budget is low though compared to traditional shoot.
crypto_nft_guy:
Feels like hype sometimes. Most AI films are just clips stitched together.
indie_dir_arun:
True, but cost reduction is real.
Earlier:
camera + crew
Now:
laptop
That changes entry barrier completely.
old_school_dir:
I’m in film industry long time.
This wont replace filmmaking, but it will change how small creators work.
Same thing happened when DSLR came.
