Idea → Finished Faceless Video
The full loop for a narration-and-visuals channel: script, voice, footage, edit, thumbnail and metadata — sized for a 3–4 hour weekly repeat.
3–5 videos / week
Total time
140 min
Total cost
$62 per run, estimated
The pipeline
- 1
Script
30 minInput: Video idea + target length
Output: Word-for-word narration script
One idea per video. Write the hook as a sentence a stranger could understand, not a teaser. Keep every claim either sourced or labelled as opinion — the script is the only place a bad idea is cheap to kill.
Prompt
Write a narration script for a faceless YouTube video, 60–90 seconds per section. Requirements: open with the strongest claim in the first 10 seconds; one idea per section; no filler words; end each section with a forward pull to the next. Flag any claim that needs a source with [SOURCE].
- 2
Voiceover
15 minInput: Final script
Output: WAV narration, loudness-normalised
Generate in 1–2 minute chunks, not one giant take — emotion flattens on long generations and a single bad line forces a full redo. Listen pass, regenerate the weak lines, export WAV, loudness-normalise before it touches the timeline.
- 3
Visuals
30 minInput: Script + narration timeline
Output: B-roll / generated shots per section
Map the script section-by-section to footage: one visual idea per script beat. Generate shots in 4–5 second units and cut them — long holds expose motion artifacts, the edit hides them. Put all text in the editor as overlays, never inside the generated frame.
- 4
Edit & assemble
40 minInput: Narration + visuals + music
Output: Cut video, beat-synced
Lay the narration first, then cut visuals to it — not the other way around. Snap transitions to the music beat, add a slow keyframed punch-in on the key line, and keep the total under the script length + 10%. Preview on a phone, not the desktop.
- 5
Thumbnail & packaging
15 minInput: Final video + title options
Output: Thumbnail + title + description
Three thumbnail concepts, each testable at phone size: if the key element is unreadable when the image is 120px wide, it fails. The title is a promise the first 30 seconds must keep — write the title before the final cut if you have to.
- 6
SEO & publish
10 minInput: Video + thumbnail + metadata
Output: Published video, tags researched
Research the search term before you write the description, not after. Chapters on (even for 8-minute videos — they change skip behaviour), and end screens that point to the next logical video in the series, not the newest one.
Keep in mind
- The voice is the brand: pick one narrator voice and commit for a quarter. Audiences recognise a voice before a face.
- Generated text fails: keep every word of copy in the editor as overlays where you can fix it.
- Cost discipline: generated visuals are the expensive step. Budget several attempts per shot and discard most — the shot list is what keeps this under control.
Total time
140 min
Total cost
$62 per run, estimated
The pipeline
- 1
Script
30 minInput: Video idea + target length
Output: Word-for-word narration script
One idea per video. Write the hook as a sentence a stranger could understand, not a teaser. Keep every claim either sourced or labelled as opinion — the script is the only place a bad idea is cheap to kill.
Prompt
Write a narration script for a faceless YouTube video, 60–90 seconds per section. Requirements: open with the strongest claim in the first 10 seconds; one idea per section; no filler words; end each section with a forward pull to the next. Flag any claim that needs a source with [SOURCE].
- 2
Voiceover
15 minInput: Final script
Output: WAV narration, loudness-normalised
Generate in 1–2 minute chunks, not one giant take — emotion flattens on long generations and a single bad line forces a full redo. Listen pass, regenerate the weak lines, export WAV, loudness-normalise before it touches the timeline.
- 3
Visuals
30 minInput: Script + narration timeline
Output: B-roll / generated shots per section
Map the script section-by-section to footage: one visual idea per script beat. Generate shots in 4–5 second units and cut them — long holds expose motion artifacts, the edit hides them. Put all text in the editor as overlays, never inside the generated frame.
- 4
Edit & assemble
40 minInput: Narration + visuals + music
Output: Cut video, beat-synced
Lay the narration first, then cut visuals to it — not the other way around. Snap transitions to the music beat, add a slow keyframed punch-in on the key line, and keep the total under the script length + 10%. Preview on a phone, not the desktop.
- 5
Thumbnail & packaging
15 minInput: Final video + title options
Output: Thumbnail + title + description
Three thumbnail concepts, each testable at phone size: if the key element is unreadable when the image is 120px wide, it fails. The title is a promise the first 30 seconds must keep — write the title before the final cut if you have to.
- 6
SEO & publish
10 minInput: Video + thumbnail + metadata
Output: Published video, tags researched
Research the search term before you write the description, not after. Chapters on (even for 8-minute videos — they change skip behaviour), and end screens that point to the next logical video in the series, not the newest one.
Workflow builder
Questions & tips
No questions yet — be the first to ask.
Loading community posts…