FaceTime-Call Ads with Grok Imagine 1.5
Two people on what looks like a screen-recorded FaceTime call, talking about the product.
Core discovery: xai/grok-imagine-video/v1.5/image-to-video speaks prompt-scripted dialogue
verbatim, with the speaker's own voice and correct lipsync — no TTS, no lipsync pass (both
lost the bake-off; a sync pass deadens the face, HeyGen avatars read as studio ads).
Reference build (copy it, don't rebuild): the marketing workspace's content/facetime-call/
— NOTES.md is the decision log, index.html the FaceTime chrome. FAL_KEY: mono/api/.env.local.
⚠️ Consent gate
Real artists' AI likenesses never ship without the artist's explicit sign-off + an AI/permission
disclosure in the post. Send each artist their clip privately for approval first (it doubles as
warm-lead outreach). Fictional characters need no gate.
Pipeline
- Script ~45-60s as alternating turns (A, B, A, B): topic-first, one proof point, organic
close ("okay, send me the link") — ship-posts' copy principles apply.
- Faces: one 9:16 FaceTime-frame image per character (below).
- ⛔ FRAME APPROVAL GATE — user OKs every frame before ANY video call. Frames cost $0.022
to fix; takes ~$2, and a frame defect multiplies into every take generated on it. Skipping
this roughly doubled the prototype's spend.
- Takes: one generation per turn. On a new frame, QC one take before batching the rest.
- Cut (every cut = character switch) → chrome (hyperframes) → ship (ship-posts).
Takes
- 15s hard cap per generation (rejected at generation time, not enqueue). Prompt in
timestamped ~5s beats:
(0-5s) … (5-10s) … (10-15s) ….
- One generation = one continuous voice; voice is NOT stable across generations. Never split
a character's consecutive lines across takes; hide every take joint behind a cut to the other
character. Alternating turns does this for free.
- Unbroken shot >15s:
extend-video (+2-10s, "what happens next" prompt) continues voice and
scene — input must be ≤15.0s (outputs run ~15.04s; trim to 14.9s first).
- Size duration to the lines (~2.5-3 words/s). Trim end-of-take dead air in assembly using
Whisper word timestamps.
Prompt grammar
FaceTime video call, subtle camera micro-shake, <room/lighting>, natural casual delivery
with clear enunciation. Exactly one person in the room, only <her> voice, <her> lips move
with every word <she> speaks starting immediately from the very first word.
(0-5s) <The character> says, speaking from the first moment: "<line>"
(5-10s) <small physical direction> and says: "<line>"
(10-15s) Looking straight into the camera <she> says: "<line>"
- No action verbs in dialogue beats — "laughs and says" produces 3s of laughing INSTEAD of
the line. Emotion goes in the tone direction; the beat verb is always plain
says.
- The one-person / only-her-voice / lips-from-first-word / one-hand-on-the-phone clauses each
kill a real artifact (phantom voices, opening mime, third hands). Keep them all.
- QC every take:
ffmpeg -vn → fal-ai/whisper (chunk_level: "word"), diff transcript
against script; pull frames at suspect timestamps. Slurs or mime-gaps → re-roll (~$2).
Faces
- Fictional:
grok-imagine-image t2i — "vertical smartphone selfie video call still frame …
grainy front-camera quality … authentic FaceTime screenshot aesthetic, not professional."
- Real person: photo →
grok-imagine-image/edit. Letterbox square sources to 9:16 first
(else x-axis squish). Prompt must say "plain raw camera image only: no user interface, no
status bar, no text" or Grok bakes a fake FaceTime UI into the image. Bars survive → one
more edit ("extend this exact scene … no black bars"). Strip gibberish clothing text if it appears.
- Reuse the same approved frame for all of a character's takes — that locks face and room.
- Artist photo sources, in order: workspace
content/*/assets/art/; og:image of their
IG post pages (profile avatars are 150px); Spotify oEmbed thumbnail_url.
Mechanics, assembly, handoff
- fal queue:
POST queue.fal.run/<model> → poll …/requests/{id}/status → GET result. Media
inputs >~700KB must go via fal storage (rest.fal.ai/storage/upload/initiate → PUT → file_url).
Validation failures cost nothing.
- Assemble: normalize takes (720x1280@30, h264, aac) → concat demuxer, trimming each at its
Whisper speech-end (+0.3s).
- Chrome: copy the reference hyperframes project (caller bar + ticking timer, rec pill, PiP of
the other character with crossfade at cuts, grain, end card; Geist fonts — the renderer can't
supply Apple system fonts).
npm run check must pass with zero errors, then npm run render.
Cost (measured 2026-07-02)
720p video bills $0.14/s (480p $0.08/s — untested quality lever); identity edits $0.022. A
debugged 50s two-character video ≈ $10; budget ~2 re-rolls per 4 takes.
1---2name: recoup-internal-video-grok-1-5-imagine-facetime3description: INTERNAL — Recoup staff tooling, gated by the recoup-internal keyword. Invoke ONLY when the request explicitly includes "recoup-internal" (e.g. "recoup-internal make a FaceTime ad"). Never use for customer-facing or artist requests. Produce a FaceTime-call-style ad — two people on a screen-recorded video call — where every talking-head clip is generated by Grok Imagine 1.5 on fal.ai speaking scripted dialogue verbatim, with native voice and lipsync (no TTS, no lipsync model). Use when the user wants a FaceTime call ad, a two-person conversation video, a Grok Imagine video, or an AI-talking-artist clip. Covers take structure around the 15s cap, the prompt grammar for word-perfect dialogue, the real-person likeness pipeline and consent gate, QC, assembly, and handoffs to hyperframes (chrome/render) and recoup-internal-social-ship-posts (publish/measure).4---5
6# FaceTime-Call Ads with Grok Imagine 1.5
7
8Two people on what looks like a screen-recorded FaceTime call, talking about the product.
9Core discovery: `xai/grok-imagine-video/v1.5/image-to-video` **speaks prompt-scripted dialogue
10verbatim, with the speaker's own voice and correct lipsync** — no TTS, no lipsync pass (both
11lost the bake-off; a sync pass deadens the face, HeyGen avatars read as studio ads).
12
13Reference build (copy it, don't rebuild): the marketing workspace's `content/facetime-call/`
14— `NOTES.md` is the decision log, `index.html` the FaceTime chrome. `FAL_KEY`: `mono/api/.env.local`.
15
16## ⚠️ Consent gate
17
18Real artists' AI likenesses **never ship without the artist's explicit sign-off** + an AI/permission
19disclosure in the post. Send each artist their clip privately for approval first (it doubles as
20warm-lead outreach). Fictional characters need no gate.
21
22## Pipeline
23
241. **Script** ~45-60s as alternating turns (A, B, A, B): topic-first, one proof point, organic
25 close ("okay, send me the link") — ship-posts' copy principles apply.
262. **Faces**: one 9:16 FaceTime-frame image per character (below).
273. **⛔ FRAME APPROVAL GATE — user OKs every frame before ANY video call.** Frames cost $0.022
28 to fix; takes ~$2, and a frame defect multiplies into every take generated on it. Skipping
29 this roughly doubled the prototype's spend.
304. **Takes**: one generation per turn. On a new frame, QC one take before batching the rest.
315. **Cut** (every cut = character switch) → **chrome** (hyperframes) → **ship** (ship-posts).
32
33## Takes
34
35- **15s hard cap per generation** (rejected at generation time, not enqueue). Prompt in
36 timestamped ~5s beats: `(0-5s) … (5-10s) … (10-15s) …`.
37- **One generation = one continuous voice; voice is NOT stable across generations.** Never split
38 a character's consecutive lines across takes; hide every take joint behind a cut to the other
39 character. Alternating turns does this for free.
40- Unbroken shot >15s: `extend-video` (+2-10s, "what happens next" prompt) continues voice and
41 scene — input must be **≤15.0s** (outputs run ~15.04s; trim to 14.9s first).
42- Size duration to the lines (~2.5-3 words/s). Trim end-of-take dead air in assembly using
43 Whisper word timestamps.
44
45## Prompt grammar
46
47```
48FaceTime video call, subtle camera micro-shake, <room/lighting>, natural casual delivery
49with clear enunciation. Exactly one person in the room, only <her> voice, <her> lips move
50with every word <she> speaks starting immediately from the very first word.
51(0-5s) <The character> says, speaking from the first moment: "<line>"
52(5-10s) <small physical direction> and says: "<line>"
53(10-15s) Looking straight into the camera <she> says: "<line>"
54```
55
56- **No action verbs in dialogue beats** — "laughs and says" produces 3s of laughing INSTEAD of
57 the line. Emotion goes in the tone direction; the beat verb is always plain `says`.
58- The one-person / only-her-voice / lips-from-first-word / one-hand-on-the-phone clauses each
59 kill a real artifact (phantom voices, opening mime, third hands). Keep them all.
60- **QC every take**: `ffmpeg -vn` → `fal-ai/whisper` (`chunk_level: "word"`), diff transcript
61 against script; pull frames at suspect timestamps. Slurs or mime-gaps → re-roll (~$2).
62
63## Faces
64
65- Fictional: `grok-imagine-image` t2i — "vertical smartphone selfie video call still frame …
66 grainy front-camera quality … authentic FaceTime screenshot aesthetic, not professional."
67- Real person: photo → `grok-imagine-image/edit`. **Letterbox square sources to 9:16 first**
68 (else x-axis squish). Prompt must say "**plain raw camera image only: no user interface, no
69 status bar, no text**" or Grok bakes a fake FaceTime UI into the image. Bars survive → one
70 more edit ("extend this exact scene … no black bars"). Strip gibberish clothing text if it appears.
71- **Reuse the same approved frame for all of a character's takes** — that locks face and room.
72- Artist photo sources, in order: workspace `content/*/assets/art/`; og:image of their
73 IG post pages (profile avatars are 150px); Spotify oEmbed `thumbnail_url`.
74
75## Mechanics, assembly, handoff
76
77- fal queue: `POST queue.fal.run/<model>` → poll `…/requests/{id}/status` → GET result. Media
78 inputs >~700KB must go via fal storage (`rest.fal.ai/storage/upload/initiate` → PUT → `file_url`).
79 Validation failures cost nothing.
80- Assemble: normalize takes (720x1280@30, h264, aac) → concat demuxer, trimming each at its
81 Whisper speech-end (+0.3s).
82- Chrome: copy the reference hyperframes project (caller bar + ticking timer, rec pill, PiP of
83 the other character with crossfade at cuts, grain, end card; Geist fonts — the renderer can't
84 supply Apple system fonts). `npm run check` must pass with zero errors, then `npm run render`.
85
86## Cost (measured 2026-07-02)
87
88720p video bills $0.14/s (480p $0.08/s — untested quality lever); identity edits $0.022. A
89debugged 50s two-character video ≈ **$10**; budget ~2 re-rolls per 4 takes.
recoupable