ugc

UGC-style means it looks like a creator filmed it on a phone: hand-held framing, single light, casual room. Imperfection beats polish. The style is allowed; deception is not. An AI avatar is a presenter for the brand, never a real customer recounting an experience that didn't happen.

1. Apply the video-hooks, video-spec-sheet, and ad-frameworks skills.
2. Define one persona: name, age range, tone, accent, look.
3. Pick the avatar (browse with get_avatar_groups / get_avatars_in_avatar_group):
   • Photoreal, off-the-shelf → HeyGen public avatar.
   • Photoreal, brand-specific → a photo avatar the user already created in their HeyGen account from a single photo (founder / employee / paid model release); no creation tool is wired.
   • Stylized / illustrated / non-human → not covered by the wired provider; say so and offer the closest HeyGen avatar.
4. Pick the voice:
   • Avatar lines → a HeyGen voice from get_voices (the render needs a HeyGen voice ID).
   • Narration over cutaways → ElevenLabs voice via search_voices.
   • Cloned founder/employee → ElevenLabs voice_clone (consent on file).
   • Localized → ElevenLabs multilingual model.
5. Before writing, name the share trigger (why viewers send it on; Instagram weighs sends over likes for non-follower reach), reveal the product in the first 3–5 s, and decide the ending: a paid ad closes on a held CTA card (video-editing, Ending row); an organic cut ends right after the payoff, loop-friendly (video-hooks, "Endings and loops"). Script (HSO for cold, PAS for problem-aware): hook and last line first, 3 hook variants, one body. The avatar speaks as a presenter ("here's what this does"), not a fictional buyer ("I tried this and lost 10 lbs"). First-person results only when voicing a real, documented customer review the user may use, with typical results stated. Save to ./video-out/<slug>/script.md.
6. Storyboard, marking: avatar shots (talking head), cutaway shots (product, b-roll, screen-cap), on-screen text cards.

Phase 2 — Avatar renders

For each hook variant + body combination, generate avatar video via avatar (generate_avatar_video with avatar_id + voice_id + script; poll get_avatar_video_status) and save to ./video-out/<slug>/avatar/<variant>-body.mp4.

If using a photo avatar: confirm it exists in the account (get_avatars_in_avatar_group) and consent is on file before the first render.

Phase 3 — B-roll cutaways

Cutaways punch up retention. ~30–40% of total runtime should be cutaways.

1. Real product footage first — clips or photos the user supplies of the product in hand and in use. Generated video still gets intricate products and hand interaction wrong, so real footage under an AI presenter beats a fully generated ad.
2. Product shots — when no real footage exists, generate_image with the user's real product photos as inputImagePaths (keeps shape, label, color); images land in ./image-out/. Set on-screen copy as text overlays, not generated pixels.
3. Lifestyle b-roll — Pexels stock first (free); generate via video-gen (Runway Gen-4 Turbo image-to-video) only when stock can't deliver.
4. Screen-cap — if the product is software, the user records once; you stitch.

Save cutaways to ./video-out/<slug>/cutaways/.

Phase 4 — Lipsync polish (optional)

If the avatar's lipsync looks soft (HeyGen sometimes drifts on consonant-heavy lines), run the cut through lipsync (Sync.so). Per-second pricing: run it only on the finished, picked variant, never on rejects.

Phase 5 — Voice + audio bed

1. Pre-render voiceover in voice (ElevenLabs) per beat, for any non-avatar narration over cutaways.
2. Generate or pick a low-key bed via music (ElevenLabs Music). UGC ads typically run muted music or none; don't fight the voice.
3. Sound design: minimal. UGC reads as authentic when it sounds like a phone-recorded clip.

Phase 6 — Stitch + variant scaling

1. Compose avatar + cutaways + audio bed with the cutaway, ducking and loudness commands from the video-editing skill's recipe reference (ffmpeg via shell). Don't improvise filter graphs or frame-processing code.
2. Use Remotion (npx [email protected] --yes --blank, npx remotion render) when generating ≥10 templated variants from the same hook + body skeleton.
3. Ship 3–5 hook variants of one body: same avatar, different first 1.5s. Hook swaps iterate one creative; ad fatigue also needs distinct concepts.

Phase 7 — Captions

Burn captions on every final cut as support: audio is primary (over 75% of Reels views are sound-on, per Meta), so voice and mix carry the ad. Use the video-captions pipeline: word timestamps of the final mix, proofread against script.md, pop preset, hook text from the script. Keep the caption block below the avatar's chin (the preset default) and inside the safe zone, never over the face.

Phase 8 — Bundle for ads platforms

./video-out/<slug>/
  ads/        variant-A-9x16.mp4, variant-B-9x16.mp4, variant-C-9x16.mp4,
              variant-A-1x1.mp4 (FB feed), variant-A-4x5.mp4 (IG feed)
  meta/       primary-text.md (per variant; front-load ~125 chars, verify per placement)
              headlines.md (≤27 chars for Facebook feed)
              disclosures.md (AI-generated content disclaimer on, any paid-creator disclosure, claims substantiation)
              landing-utms.md (UTM-tagged URLs per variant)
  ads-library-checklist.md

Skills

• video-hooks: always; UGC lives or dies in the hook.
• video-editing: Phase 6 stitch, cutaway and audio recipes.
• video-captions: Phase 7, every final cut.
• video-spec-sheet: Phases 6–8.
• ad-frameworks: Phase 1; HSO default, PAS for problem-aware.
• content-voice: when the persona needs a defined tone.
• content-humanize: always; strip AI cadence from VO scripts.

Memory protocol

Before starting: remember(["persona", "avatar id", "voice id", "winning hooks", "banned claims", "consent on file"]). After completing: memorize() persona that worked, hook patterns by vertical, lipsync polish needed (yes/no), avatar/voice combo wins.

• Avatar in frame for most of the runtime; cutaways support it rather than replace it.
• First 1.5s is the avatar speaking the hook, not a logo or stock cutaway.
• Single creator across all variants in a campaign; mixing creators reads as branded.
• No music duck on the avatar's voice. Music should be -18 dB or off.
• Captions inside the safe zone: Meta Reels keeps the top 14%, bottom 35% and 6% sides clear (see video-spec-sheet).
• Length: 15–30s is the usual starting range for paid UGC; let the account's hook and hold rates decide. Both platforms accept far longer ads.
• Test hook variants in Meta Ads Manager, one ad set per hook.

Compliance (2026 — verify live)

• FTC rule on fake reviews and testimonials (16 CFR 465, in force since Oct 2024): bans testimonials that misrepresent who the speaker is or what they experienced, AI-generated ones included. The Endorsement Guides treat virtual influencers as endorsers.
• TikTok Ads Manager's AI-generated content disclaimer is mandatory for AI avatars and synthetic voices; undisclosed ads are rejected or restricted. Meta labels detected AI content "AI info".
• TikTok commercial-content disclosure toggle on for paid or branded posts; undisclosed posts can lose For You eligibility after 24 h.
• EU audiences: AI Act Art. 50 requires deepfake disclosure from 2 Aug 2026.
• New York: ads featuring a synthetic performer (an AI-generated person who isn't a recognizable real individual) need a conspicuous disclosure (GBL § 396-b, in force since 9 Jun 2026). Disclose stock AI avatars rather than argue the edge; note it in meta/disclosures.md.
• A real person's likeness or voice (photo avatar, voice_clone) needs their written consent on file.

Do:

• Keep the avatar a presenter, with any real-customer quote documented and permissioned.
• Confirm consent before rendering with any real person's photo avatar.
• remember() before, memorize() after.