Why are most GTA clips on Twitch Japanese?
8 min read

Because Japan’s Grand Theft Auto roleplay scene clips constantly, and the audience that actually dominates GTA viewing is not on Twitch at all.
Of the 320 most-viewed GTA clips we hold from Twitch, 293 are Japanese-language, and they carry 93.9% of the views — 1,511,631 of 1,609,843. Fifty-nine separate Japanese channels appear in that set, out of 81 channels in total. This is not one big streamer. It is a scene.
The single most-viewed clip in the whole set is English, and it makes the rest
sharper rather than softer: DarkViperAU’s discord subpoenaed, 26,536 views.
One English clip beat every Japanese clip individually. Collectively the
Japanese ones beat it twenty-six times over.
The two measurements disagree, and that is the finding
Clip views are not watch time, so the honest comparison is against Twitch’s own numbers rather than the whole category. On our board of the 100 most-watched GTA creators, the Twitch rows carry 5,172,269 hours watched over the current 30-day window, and Japanese-language channels are 8.0% of that.
| Language | Share of Twitch GTA clip views | Share of Twitch GTA hours watched |
|---|---|---|
| Japanese | 93.9% | 8.0% |
| English | 3.6% | 46.8% |
| French | 0.0% | 32.3% |
| Italian | 0.0% | 7.6% |
| Russian | 1.0% | 4.0% |
| Everything else | 1.5% | 1.4% |
French is the row that gives the game away. French-language channels are a third of Twitch’s GTA watch time on that board and produced 200 clip views in the entire corpus. A large, healthy audience that does not clip looks, in a clip ranking, like nothing at all.
The result holds however you slice it. Restrict to clips created in the last seven days and Japanese is 96.7%. Take only the top 50 clips by views and it is 90.4%. Drop the largest Japanese channel entirely and it is 93.4%. It is not one viral moment and it is not one creator.
What they are clipping
Almost all of it is one thing: ストグラ — Sutogura, short for Street Graffiti Roleplay — a Japanese GTA V roleplay project where streamers play residents of a shared Los Santos with no script. Season 1 ran roughly 600 broadcasters and around 800 characters, and Season 2 began on 19 December 2025 (reported — RealSound, 5 January 2026).
Read the clip titles and it is a soap opera. 隊長の降格報告と新隊長 就任の挨拶に行ったら、警察からヤバすぎる情報漏洩の話でカウンターを喰らう救急隊 — the EMS team goes to announce a captain’s demotion and gets blindsided by news of a police information leak. 警察は常時 情報漏洩 中だった…? — “the police were leaking the whole time…?” 14名以上の住人が街から去る日 — the day more than fourteen residents left the city. These are plot beats, clipped and passed around, from a continuing story with a cast.
That is a different reason to clip than the one the English-speaking clip economy runs on. A funny death or a good shot is self-contained. A plot revelation only means something if you follow the show — and if you follow the show, you need the clip to keep up with the episodes you missed.
The scale is not something we discovered.
Stream Hatchet reported in April 2026
that across the twelve months to February 2026, GTA V “leads by a significant
margin at 230M hours watched” among Japanese audiences, naming Junichi Kato,
Sakura Miko and Kuzuha as drivers (reported). The top clipped channel in our
own set, 叶ちゃんねる, is the NIJISANJI streamer Kanae, who posts ストグラ
broadcasts regularly (confirmed from his own X account);
on our creator board he sits at rank 20 worldwide with 24,665 average viewers
and a 36,025 peak.
Why our own boards never showed us this
Here is the part that is a criticism of our method rather than a finding about Japan, and it is the more useful half.
We take our detailed censuses at around 04:00 UTC. That is 13:00 in Tokyo — the middle of a Japanese working day, and the deepest part of the trough for Japanese streaming. In a complete census at 04:26:08 UTC on 1 September 2026 — board plus tail, 1,925 channels reconciling exactly against our published total of 166,872 viewers — there were nine Japanese GTA streams live, holding 49 viewers between them. The biggest had sixteen.
| Measure, GTA on Twitch | Japanese share |
|---|---|
| Most-viewed clips, 29 Jul – 1 Sep | 93.9% |
| Hours watched, top-100 board, 30 days | 8.0% |
| Live viewers, full census 04:26 UTC 1 Sep | 0.03% |
Three numbers, same game, same platform, spanning four orders of magnitude. The snapshot is not wrong — at that minute there really were 49 people watching Japanese GTA. It is just answering a question about one minute of the day and being read as a question about a scene. We have written before that GTA’s audience swings nearly four to one across the UTC day; this is what that swing does to a population rather than a total, and it is a stronger argument than we had for sampling twice a day instead of once.
What this does not show
Clip views are counted early, and that flatters whoever clips fastest. We ask Twitch for the category’s top clips from the preceding two days, so a clip’s stored view count is roughly its first 48 hours. A clip that keeps accruing views for a month is undercounted here. If Japanese clips are shared onto X within hours and English GTA clips build slowly, some of the gap is that and not behaviour. It cannot be most of it — a twelve-fold gap does not come from a counting lag — but it is real and we cannot size it.
This is Twitch only. We collect no clips from Kick or YouTube, and Kick is where 83.3% of the watch time on our creator board sits. So this measures the clip culture of the platform that is now a minority venue for this game, and says nothing about whether Arabic or Portuguese GTA on Kick clips heavily. It probably does; we cannot see it.
The 8.0% is a floor. Only 23 of our top 100 creators are on Twitch and exactly one of them is Japanese, so that figure rests on a single channel. Japanese creators below the global top 100 are not counted in it, and there are plainly many. The true Japanese share of Twitch GTA watch time is higher than 8.0%, which shrinks the gap by an unknown amount.
Language is self-declared, taken from the channel’s own Twitch setting, and the clip inherits it. And a clip view is not a person: it counts plays, we cannot deduplicate them, and it says nothing about attention.
None of that touches the ordering. Whatever the exact multiple, the language that produces almost all of Twitch’s most-watched GTA clips is not the language that watches the most GTA on Twitch, and it is nowhere near the one that watches the most GTA anywhere.
Sources, and how to check us
The clip figures come from data.luciastream.com/boards/clips-trending.json, as
published at 03:54:41 UTC on 1 September 2026: the 320 most-viewed GTA clips
we hold from Twitch, created between 29 July 2026 and 01:16 UTC on 1 September
2026. Watch-time figures are the Twitch rows of
boards/top-creators-7d.json, published in the same pass — a 30-day rolling aggregate
regardless of the filename, which is
a thing we got wrong in print for three weeks
and corrected today.
The live figures come from 18 complete censuses between 04:09:10 and 04:25:09
UTC on 1 September 2026 — mean 185,359 viewers on 1,939 channels, per-platform
sums reconciling exactly against live/summary.json in all 18, zero duplicate
(platform, stream) pairs, no platform reporting degraded coverage — plus the
single full board-and-tail census at 04:26:08 UTC quoted above.
If you want to quote any of it, the permalink is
https://luciastream.com/blog/why-are-gta-clips-japanese/, the sampling windows
are the ones named above, and the collection method is written up at
luciastream.com/methodology. The live
numbers refresh every 60 seconds on our dashboard,
the creator ranking is on the leaderboard,
and the clips board is the same data this post
is built from, so you can check whether it still reads this way by the time you
get there.
Numbers in this post come from our own sampling — seehow we measure.