-
A replica is a new performance.
GamesIndustry.biz reported
on 20 October that the
Japan Actors Union has
announced guidelines on
the unauthorised use of
generative AI to replicate
voice performances. The
piece, citing Automaton,
says the framework
recommends prior consent
at every stage, including
before a model is trained
on an actor's performance.
If generative AI replicates
a performance, the union
says treat it as a new
performance. That means
additional royalties or
compensation for the
actors involved. A file
that sounds like the take
is not the take. It is
another use, and another
use has to be paid for.
Deceased actors are
narrower still. The
guidance says a voice may
be replicated only if
permission was granted
during the person's
lifetime. Developers or AI
companies must also obtain
consent from the family
and from the affiliated
agency or company, and pay
royalties. A famous bark
in a folder is not a
lifetime yes.
Automatic translation sits
in the same note. If
generative AI translates a
performance into another
language, it must keep the
artistic integrity of the
original, and it cannot be
used overseas without
permission. A store-page
localization is a stage.
It is not a free side
effect of the first
recording.
The union said it will try
to make the guidelines
industry-wide rules, and
it will push for laws that
protect what it calls
voice rights. These are
recommendations, not a
statute you can hide
behind. Until someone can
name the stage and the
yes, the line does not
ship.
Source:
GamesIndustry.biz — Japan Actors Union voice guidelines
-
The mark is not the list.
GamesIndustry.biz reported
on 19 October that UKIE
has partnered with law
firm Sheridans on an AI
Compliance and
Accreditation Service for
UK studios. The service is
a four-step process that
ends in a UKIE
Certification Mark. The
mark is meant for pitches
and funding rounds, as a
way to show compliance and
due diligence.
UKIE CEO Nick Poole told
the site the aim is a
practical route through a
fast-moving area, and a
standard the wider
industry can recognise. He
said most members are
already experimenting with
these tools in different
ways, and the shared worry
is knowing where the legal
and commercial risks sit,
then being able to show
they took those risks
seriously. A publisher
wants a considered
approach. A vibe is not
that.
Sheridans partner Anna
Poulter-Jones said the law
and public policy keep
moving, so a studio policy
should stay under review.
The service can look at a
policy that already exists,
or write one from scratch,
and a studio can come back
when the law moves or when
the way they use AI
changes in a substantial
way. A policy from last
year is not a policy for
the build you are about
to upload.
The article gives prices.
A compliance check plus a
new AI policy is typically
£1,500 to £2,000 for UKIE
members, and £2,000 to
£2,500 for non-members. A
review of an existing
policy plus an ongoing
check is £500 to £750 for
members, and £750 to
£1,000 for non-members.
Poole said the look covers
the tools in use, the
licences, whether
AI-generated material
reaches players, and the
processes already in
place.
A mark on a deck is a
later signal. It does not
replace the list. Before a
generated file reaches a
player, write the tools,
the licences, and whether
that file is player-facing.
If you cannot name those
three, you do not have a
considered approach. You
have a badge-shaped wish.
Source:
GamesIndustry.biz — UKIE and Sheridans accreditation
·
UKIE — certification mark
-
The seam is not the song.
SEGAN 0.4.0, dated 20
October, adds soundtrack
families, bar-aligned
crossfades, and Ogg Vorbis
export for music. The tool
is open source. Agents can
write instruments, songs,
and sound effects as text,
then check and render them
locally. Sound effects stay
WAV. Music, variants,
stems, and flow clips can
go out as Ogg if the
optional Ogg dependency is
installed.
Families are the join rule
for related songs. When a
game state picks another
song in the same family, it
enters on a bar boundary
and overlaps the previous
song. Songs from other
families keep the fade-out,
the configured gap, and
the restart. An exploration
area can group a few
candidate loops into one
family and give that state
a crossfade duration.
Selection still belongs to
the music director. The
update does not decide
which loop belongs in the
cave.
Stem mixes get a trim
adjustment so they can hit
a loudness target in
playback. A limiter ceiling
caps that adjustment, and
a warning flags mixes that
still cannot reach the
target. A warning is a
stop. It is not a
suggestion to export
anyway. Godot 4 gets
companion resources, and a
project with a flow song
or a soundtrack also gets
a music-director script.
Other engines have to
integrate the exported
data themselves.
The project installs from
a GitHub version tag, needs
Python 3.11 or later, and
has no PyPI release yet.
The code is MIT. During
the 0.x series, minor
versions may change formats
or exports, so pin a tag.
Listen sheets already show
notes, waveforms,
spectrograms, and loudness.
Those are evidence for
timing, clipping, and loop
seams. They cannot decide
whether the music suits
the game. Assets stay
drafts until a person
listens and approves them.
Pin v0.4.0 if that is the
export you mean. Then
listen once, on the loop
the player hears. A clean
seam is not a yes.
Source:
MakeGameWithAI — SEGAN 0.4
·
SEGAN v0.4.0 changelog