RECURRING CHARACTER · OWN CHANNEL · 9:16 · EN 0:59 · JA 1:05 · 2026Field notes
Recurring character ad,in English and Japanese
A generated host with a fixed voice who sells a real product, in English and Japanese, for Instagram, TikTok and Shorts.
REM Dealer is our own character channel; the product is a real store listing, sold with an affiliate link. Not made for, or paid by, the brand. REM is a generated character with a synthetic voice.
remdealer-en.mp4 · 1080×1920 · 30 fps · 0:59 · −14.5 LUFS integrated, measured on the delivered file
FIELD NOTES
How it was made
One character, one fixed voice.
REM is a generated character with a synthetic voice, the character's own fixed voice. He is generated in every shot where he appears, and the voice was checked against the character's reference voice: every word starts within 21 ms in English.
Two languages, eleven sections in the same order.
The English piece runs 0:59 and the Japanese 1:05, both with eleven sections in the same order, so the player can cut from one to the other at the start of the same section. The on-screen text was reflowed per language and kept inside the safe zone: measured text from 232 to 1289 px in English and 236 to 1287 px in Japanese, inside 220 to 1436 px.
A real product, declared.
The product is a real store listing, sold with an affiliate link. Its page was captured by code, without price, stars or reviews, and the affiliate disclosure is said, captioned and written in both languages.
Measured, with two checks that didn't pass.
Captions, on-screen text, the mix and the measurements were built and measured in FireNgine, our production system. Loudness at −14.5 and −14.4 LUFS, true peak at −4.9 and −5.0 dBTP, worst text contrast 11.2:1 and 10.9:1, voice level across segments 2.2 and 1.7 LU. The Japanese version was transcribed back and checked against the script. The end of "c'mere" lands 60 ms late against a 40 ms limit, and one Japanese word ends 100 ms late: both delivered with their numbers.
Numbers
| check | English | Japanese |
|---|---|---|
| loudness (target −14 LUFS ±1) | −14.5 LUFS | −14.4 LUFS |
| true peak (max −1 dBTP) | −4.9 dBTP | −5.0 dBTP |
| text inside the safe zone | yes | yes |
| worst text contrast (min 4.5) | 11.2:1 | 10.9:1 |
| voice level across segments (max 6 LU) | 2.2 LU | 1.7 LU |
| affiliate disclosure | said, captioned and written | said, captioned and written |
| the character's voice | didn't passevery word starts within 21 ms, but the end of "c'mere" lands 60 ms late against a 40 ms limit | didn't passthe whispered hook sits slightly off the character's reference voice, and one word ends 100 ms late against 40 ms |
The Japanese version is transcribed back and checked against the script.
Provenance
Generated: the character, his voice and the shots he appears in. The rest is licensed or made in FireNgine.
Have a script, a voice-overor a product to launch?
Send a short brief or write to us. We'll come back with questions, then one central idea for the piece before anything is made.
hello@firendev.com