Video Ad Standards

• ✓Max 2–3 words on screen at once
• ✓One line only — split anything that wraps to two lines
• ✓Replace every period with a triple dot ... — periods feel too final
• ✓Font: Montserrat anywhere there's a font choice
• ✓Default colors: red background, white font
• ✓Test variants allowed: green, yellow only
• ✕More than 3 words on screen
• ✕Two-line subtitle blocks
• ✕Periods at end of subtitle phrases
• ✕Any font other than Montserrat without strategist approval

• ✕Text-to-speech (TTS) AI voices in any creative — the prosody is the giveaway, not the timbre
• ✕Female TTS voices specifically — worst uncanny-valley issue
• ✕Obviously-AI visuals in the hook
• ✓Voice conversion (speech-to-speech) is allowed and preferred — see Voice Production
• —AI visuals outside the hook: testable but discouraged

The default voiceover approach is voice conversion (speech-to-speech), not TTS. A human records the script with the right pacing, emphasis, and emotion; an AI tool swaps the timbre to a target voice. Prosody stays human, so the "AI giveaway" disappears.

Default tool: ElevenLabs Voice Changer (Speech-to-Speech)

• ✓Record the script with intended pacing, emphasis, breaths, and beats — phone mic is acceptable
• ✓Run through ElevenLabs Voice Changer with the target voice selected
• ✓Use a consistent target voice per niche / per client where possible
• ✕Generating voiceover from text (TTS) — even with "natural" voice settings
• ✕Skipping the human source recording

• —Only for ads where ElevenLabs prosody isn't surviving the swap
• —Higher cost, slower turnaround, Hollywood-grade emotional nuance
• —Not the default — most ads do not need this tier

• ✓Look like the contractor recorded it on their phone — UGC feel
• ✓Subtitles + color are the only "produced" elements
• ✓One project per video — feels like one job site, not a montage
• ✓Speed-match every clip to the rest of the video (e.g. 5x a slow clip if the rest runs fast)
• ✕Polished agency look
• ✕Mid-video pacing breaks (slow-mo dropped into fast-cut sequences)

• 1.Open with a before/after when client footage allows
• 2.Middle = process (drilling, wrapping, pouring) in correct chronological order
• 3.End on the finished result — fully wrapped, fully poured, fully encapsulated

• ✕Out-of-order process clips (wrap → dig → wrap → dig)
• ✕Ending on mid-process footage

• ✓Use client-provided before/afters first — highest-trust footage
• ✓If client provided afters only, pair with generic before footage
• ✓Every visual must match the script category exactly
• ✕Fabricating a before
• ✕Showing epoxy in a decorative-concrete ad (or any category mismatch)

• ✓Lead crawlspace repair as the primary frame
• ✓Incorporate basement waterproofing as a secondary element

• ✓Visuals must show decorative concrete — not epoxy, not stamped, not poured-and-broomed
• ✓Male voiceover preferred over female AI

• ✓Subtitles ≤3 words, one line, no periods (triple dots only)
• ✓Font is Montserrat
• ✓Color scheme is red/white (or approved test variant)
• ✓No AI voice detectable in first 3 seconds
• ✓No AI visuals in hook
• ✓Reads as phone-shot UGC
• ✓All footage is one project (or feels like one)
• ✓All clips are speed-matched
• ✓Open is before/after when client footage allows
• ✓Process clips are in chronological order
• ✓Final shot is the finished result
• ✓Every visual matches the script category
• ✓Crawlspace ads lead with crawlspace

• —Crawlspace: opening on before/after; real person on camera (sped up to match)
• —Decorative concrete: male actor voiceover; decorative-concrete b-roll matched to decorative-concrete script

• —Crawlspace: AI-feeling hook; out-of-order wrap/dig sequence; slow-mo clip in a fast-cut video
• —Decorative concrete: female AI voiceover; epoxy footage shown over a decorative-concrete script

• —Color variants beyond red/white (green, yellow open for A/B)
• —AI usage outside the hook
• —Generic before-footage sourcing standard when client provides afters only