[Sonomama] Mushoku Tensei: Jobless Reincarnation - S03E13v2 (1080p WEB AV1 Dual-Audio DDP2.0)

Spotted a problem? Please leave a comment on this page, whether it's a typo, a mistranslation, a timing issue or anything else. Every fix reported here will be applied in the season batch release.

This is the first release on our new AV1 encode. The weekly files use one CRF for the whole episode (CRF 18) and a fixed grain level. This file is encoded scene by scene instead: every scene gets the cheapest CRF that still meets an SSIMULACRA2 target, measured on every frame against the source, and the source's own grain is measured and synthesized back 1:1. Graded the same way across the season, none of the 14 weekly encodes met that target, and all 14 new ones do. The full settings are in the table below. The season batch brings the same encode to every episode.

1080p AV1 dual-audio release: our own AV1 encode of the VARYG CR WEB-DL video, with Amazon's Japanese and English audio from AnoZu.

Video	VARYG (Crunchyroll WEB-DL, H.264 1080p) is the source. The picture is our own CPU AV1 encode of it: 10-bit (yuv420p10le), 1920x1080, 23.976 fps
Audio	Japanese + English, both Amazon's delivery of the same mixes, from AnoZu: E-AC-3 2.0 224 kb/s, 48 kHz, untouched (Japanese default). Chosen over Crunchyroll's AAC after measuring both on E13: same mix and timing, but Amazon's encode keeps the audio band to 20.4 kHz where Crunchyroll's AAC rolls off at 17 kHz
Subtitles	[Sonomama] (ours, default) · AT-X CC · DMM TV CC · Crunchyroll · Signs · SDH [CR] · Simplified / Traditional Chinese · AR, DE, ES-419, ES, FR, ID, IT, MS, PL, PT-BR, RU, TH, VI — official tracks untouched, fonts attached
Target SSIMULACRA2 (per scene, every frame, vs the source)	scene mean ≥ 82, 5th percentile ≥ 78, minimum ≥ 70 — each scene gets the cheapest CRF that meets all three, and the encoder is never asked to go below the source
Encoder	SVT-AV1 5fish v2.3.260827: --crf 14–40 per scene (this episode's 276 scenes: 59 at 14, 7 at 17, 162 at 20, 27 at 28, 11 at 34, 10 at 40) --preset 2 --lp 6, variance boost on (strength 2, octile 5), --film-grain-denoise 0, BT.709 limited range, a forced keyframe on every scene cut — identical bit-exact output from four machines
FGS	film grain synthesis tables rewritten per scene to match the source's own grain 1:1 (measured, not a fixed level) — the coded picture underneath is untouched
Filters	none — no denoise, no dehalo, no debanding. The source picture is encoded as-is and its grain is synthesized back rather than filtered away

Subtitles: the default English track is our [Sonomama] edit, with no song lyrics this week - this episode has neither an opening nor an ending sequence. The untouched official Crunchyroll tracks are also included (English full/Signs/SDH plus every other language CR shipped for this episode), plus two Japanese CC tracks: AT-X CC (the broadcast ARIB captions, extracted from [shincaps]' AT-X TS captures, CMs and sponsor cards removed, re-timed to this video) and DMM TV CC (sourced from Jimaku, re-timed to this video). All [Sonomama] scripts are authored at 1920x1080. Fonts attached.

About the [Sonomama] track and the MTL tag: the base is Crunchyroll's official human translation. We check it line-by-line against the Japanese subtitles (broadcast ARIB captions and the JP CC tracks) AND against the original Japanese light novel (the raw JP text, not a translation) - the novel is our ground truth for names, terminology, intent and each character's speech register, not just the JP captions. We restore honorifics, remove localizer-invented lines, and correct mistranslations. Where a line is corrected it is re-rendered from the Japanese with machine assistance, and every edit is reviewed by two independent auditors before it is applied. Because machine translation is in the lineage of those corrected lines, the MTL tag is set. If you want the untouched official script, it is included as a separate track.

Sub work on this release: translation check + editing of every line against the JP captions and the raw JP light novel; timing corrections where the JP audio/captions show CR's cue early or late; and every episode is watched in full and verified before it ships.

Video and the official subtitle tracks are sourced from VARYG, and both audio tracks (Amazon WEB-DL) from AnoZu - both credited as secondary groups.

File names are based on TMDB with target compatibility to Jellyfin. You may still need to rename for your own library organization.

What changed vs the first release [386AA121]

The original E13 went out Japanese audio only on Tuesday, September 22. The English dub and the DMM TV
caption capture were not out yet at that point, so it could not carry them. This v2 folds them in, and it is the
first file on our new encode:

• New video encode — the per-scene AV1 encode described in the table above, replacing the flat CRF 18 encode.
• Audio from Amazon, both languages — Japanese and English E-AC-3 2.0 224 kb/s from AnoZu's Amazon WEB-DL,
  replacing v1's Crunchyroll AAC Japanese track. It is the same Japanese mix and the same English dub Crunchyroll
  carries, checked in sync with Crunchyroll's tracks (0 ms offset). English stays non-default.
• Crunchyroll Signs and SDH [CR] subtitle tracks added.
• DMM TV CC Japanese caption track added, below the existing AT-X CC.
• One [Sonomama] correction: 00:13 — Rudeus's "Yeah. It gave me the creeps." → "Mm. It gave me the
  creeps." The Japanese is a hum (うん), and a hum inside a line is now spelled as the sound across the season.

Nothing else changed. The AT-X CC, the Chinese pair, the full Crunchyroll English track and every other
language [386AA121] carried are byte-identical here — verified per stream, not per file.

21 subtitle tracks total: EN (Sonomama/CR/Signs/SDH) + JP CC (AT-X/DMM TV) + ZH-Hans/Hant +
AR/DE/ES-419/ES/FR/ID/IT/MS/PL/PT-BR/RU/TH/VI.