3.0.0b29 -- a run that falls short says so, and stops

Added

• The check before a run names the Auphonic preset's processing
  switches in words: each one that is on, with its setting, and that
  the program changes none of them.
• A second run of a single recording now finds its Auphonic production
  again, as Multitrack already did, and asks: take the result, upload
  anew, or stop. --auphonic-resume answers for a script, and --help
  no longer marks it as Multitrack only.
• The Auphonic box now says before the run whether the credit is
  enough for what auphonic.com will charge, each production counted at
  three minutes or more, as the price list says.
• The Python floor and the tested package versions are now one
  command away: the README gives one install line per system with the
  versions the release's tests ran green on; the pipx line does the
  same.
• The check before a run names an intro or outro stored in the
  Auphonic preset as not used, for Multitrack and Singletrack alike:
  intro and outro come from the video files set to Intro or Outro.
• Two voices of one speaker separation that sound alike are named:
  the log says so in one line, also in a run the window hands its
  separation to, and the row's Separated turns amber in the window.
• The check before a run counts clipping in a camera's own sound
  where that sound is a track of the production, and in no other
  camera's.

Changed

• The program now needs Python 3.12 or newer; 3.10 and 3.11 are
  refused at start and by pip. The package versions a release is tested
  with no longer install below 3.12.
• The "Multitrack" tick and --multitrack are gone. The production
  kind follows from the material, and a grey line under the preset list
  says which kind and why. A command line with --multitrack is refused.
• A preset that lets auphonic.com cut silence, fillers, coughs or
  music now stops the run before the upload, because the sound would
  come back shorter than the pictures. The advice names the place in the
  preset; "Set Cuts To Silent" keeps the length and goes on.
• Transcripts, subtitles and speech data are no longer fetched from
  auphonic.com, and no request asks for speech recognition any more,
  Multitrack or Singletrack. The manual's promise "no text from
  auphonic.com" now holds.
• A file of more than two channels that would go to auphonic.com now
  stops the run before anything is sent, naming the file and its
  channel count: auphonic.com takes at most two channels in one file.
  The channel rows in the window cut such a file into tracks first.
• On the free auphonic.com plan, a Multitrack production longer than
  about 21 minutes now stops the run, in the check before the run and
  again before the upload, with the ways out named. --anyway lifts only
  the first stop, and the last line says so; a dry run only warns.
• A cut Timeline that lacks something now ends the run with errors:
  no picture, fewer shots than were sent, no Full-Mix, a refused intro or
  outro, or an empty or silent cut. The window says "Finished with
  errors." and the end of the log lists what is missing.
• A fault in the program's own code at a step in Resolve now stops the
  run and says it is a fault of the program, to be reported, with the
  details in the log. It used to pass for a refusal by Resolve.
• A single recording sent to auphonic.com is now made from the preset's
  settings without its intro and outro, the way a Multitrack production
  is made.
• A dry run no longer writes the tracks onto the time axis for its own
  use: a 68-minute interview needs about 2.4 GB less room, and the cut
  stays the same. With an API Key, or with microphones that hear each
  other too well, the tracks are still written.
• Without a picture, the window and a run measure the time axis only
  once: each takes what the other measured over the same recordings, as
  they already did beside a picture.
• A camera that gives no sound at all no longer stops the run when no
  other camera is heard: the recordings are laid against each other as
  without a picture, and each camera stands at its timecode. The window
  lays every file at the same place.
• A recording without a picture whose level is adjusted is now always
  written at 48 kHz, as the check before the run announces.
• Pauses inside a turn are now bridged up to 0.75 s instead of 0.35 s
  where the program reads from the tracks who speaks: on real interviews
  the camera stood on the speaker more often, 92.7 instead of 88.5 % of
  the time, with as many cuts.
• The transcript line in the log now gives only the number of words;
  its three shares followed how right the names were only weakly.
• "Wide shot at the latest" now says what it does: the clock breaks
  up a shot only when it is longer than both this and "Wide shot after",
  and the greeting and farewell wide shot ends there either way; the
  manual and --help say the same. The cut is unchanged.
• --update now installs a release with that release's own list of
  package versions, so going back to an earlier version brings that
  version's packages too. Releases before 3.0.0b28 carry no list and are
  installed without one.
• ffmpeg is now fetched from one named build and checked against a
  checksum kept in the program, no longer from whatever the build host
  calls "latest". A newer ffmpeg comes only with a newer program.
• On a Mac, recognition now hears a language in one fixed variant:
  the one asked for, the Mac's own region (de-CH on a Swiss Mac), then a
  fixed one per language such as en-US. English on a German Mac used to
  land on a different variant run after run.
• A language the Mac has no recogniser for now goes to Whisper, and
  the log says so, for example "macOS hears no Dutch here, Whisper takes
  it". Before, the file was heard in the system language.
• A recording in blocks is now heard once for the window's
  transcript: the window takes the words of the voice separation
  instead of listening a second time.
• The project file now keeps the reference camera, so a reopened
  project reads a typed timecode mark at the rate of the camera the run
  used.
• German now says "Transkript" everywhere, in the window and the
  manual, no longer also "Niederschrift".

Fixed

• A voice name typed just before the assignment sheet is rebuilt now
  stays. When a file's tracks finished cutting in the background at
  that moment, the name silently went back to "Speaker N".
• Quitting the window no longer hangs. Stopping the players at quit
  could deadlock inside Qt's media backend, about one quit in 75 under
  load; the players are now left to the system as the program ends.
• The window no longer hangs while playing: a deadlock between
  PySide 6.12 and Qt's media thread is now avoided when each sound output
  is made. It is reported to Qt as QTBUG-151644.
• The window now opens on a Qt without its multimedia modules, and no
  longer opens ffplay by itself there; a file handed to the player still
  goes to ffplay.
• The window now lays a recording's later blocks where the run puts
  them. After a gap in the timecode the preview played a block 1.5 s
  too early; it now matches the run to the millisecond.
• A Multitrack preset that starts with an intro now sends the speaker
  tracks with the preset's track settings, no longer with the intro's.
• Each Auphonic production lookup now takes only its own kind: a
  Multitrack run no longer finds a Singletrack production of its title.
• A second run now finds its production at auphonic.com even when more
  than ten newer ones lie before it: the list is read back to the
  newest 50. Before, a production further back was uploaded and paid for
  again.
• A command-line run now goes up to auphonic.com under its
  production's name, from --production or else the folder, so this
  week's run no longer finds last week's production titled "Production".
• A start auphonic.com refuses, or a production it holds as
  "Incomplete", now stops the run at once with auphonic.com's answer,
  instead of waiting until the time limit, two hours by default.
• A refused upload of a Multitrack track now stops before the production
  is started, and a preset read that comes back without a preset stops
  before anything is created.
• A returned Auphonic track that lacks part of what went up is now
  refused: its line says "refused" and how many seconds are missing, and
  the run stops. Before, a return a few seconds shorter than what went
  up passed as aligned.
• With --auphonic-done, a file trimmed to a later window is now judged
  as it lies in the folder; padded with silence, a file up to a minute
  short at its end used to pass the return check.
• On the free plan, the check before a run now counts tracks as the run
  sends them: a recording set to "do not use" counts as none, two
  recordings of one name as one, a camera whose sound is used as one.
• Microphones joined into one recording are now separated for one
  person per microphone, unless --speakers-count gives a number: one
  voice too many used to leave every voice without its microphone, 58
  instead of 95 % of the cut right.
• A Sync run whose second recording is set to "do not use" is no longer
  stopped by its own check, and in the window no stop suggests
  --anyway, which the window cannot pass.
• The check before a run leaves a recording set to "do not use" out of
  the disk space and the loudness line, and a finding about a file that
  takes no part no longer stops the run.
• German and Greek no longer put two full stops after a length ("30
  Min..") in the free-plan sentence and the credit line.
• A whole run with --resolve whose Resolve part stops now ends with
  errors: the window says "Finished with errors." instead of "Done.",
  and the command line returns 1.
• **The...

3.0.0b28 -- the preview stops rehearsing and simply does the run

Added

• A stereo mix finished elsewhere can now stand in for the Full-Mix:
  set "In the sound" to "Finished mix", or give --finished-mix FILE.
  It is placed on the time axis and goes into the camera files and the
  Resolve handover as it came; it is no speaker and is not sent to
  auphonic.com.
• The Auphonic box now shows the account's plan and the minutes of
  credit left as soon as the account answers, also without a preset
  and with "work without Auphonic". The line turns red when the
  recordings need more minutes than are left.
• The check before a run now asks the Auphonic account once and says
  whether the minutes left carry the production. On the free plan a
  Multitrack production longer than 21 minutes gets a warning, there and
  again before the upload; the run still goes on.
• Every release now carries one list per system of the package
  versions its suite ran green on (constraints-macos.txt,
  -windows.txt, -linux.txt); pip3 install -c constraints-macos.txt ... installs exactly those.

Changed

• The cut preview is now the run itself: a dry run of what the
  window holds, stopped before it writes, in the background and without
  a line in the log. Preview, "Dry run" and "Start" cut the same shots.
• A changed cut value now works out only the cut again. A changed
  file, name, camera, In point or Out point runs the preview again over
  what is already measured.
• "Start", "Dry run" and "Create Resolve project" no longer wait for a
  preview still working: they stop its run and begin at once.
• After "Start", the cut preview shows the finished run's handover
  and says "from the finished run". A run with other settings does not
  replace it; a later dry run of the same settings takes over again.
• Once the window has written its transcript, the preview and the run
  cut by the same words: a question that becomes a reaction cut in the
  preview is one in the run too.
• "Dry run" now works everything out up to the finished cut and the
  Resolve handover, and still writes nothing. The run after it takes
  the time axis and who speaks when from what the dry run kept.
• Window and run now share one time axis: whichever measures first,
  the other takes it without measuring again. A recording in blocks is
  measured joined in the window too, as the run measures it.
• Cameras now line up the same whatever order they are given in. Of
  cameras as long, the reference is the one placing most others, then
  one with a timecode, then the first by name.
• A camera the reference camera does not hear is now placed through
  one that hears it, and its line in the log says through which. A
  recording that shares only its last minute with a camera is found.
• The log now names a camera whose clock and measured place are more
  than a frame apart, with the gap; the measurement is used. Without a
  timecode on the reference camera, every written camera carries the
  timecode of its measured place, from 00:00:00.
• A person assigned to one camera by their track and to another by
  their voice now sits on the voice's camera in the cut list, the
  handover and the preview alike, and the log says so.
• If building the Resolve project stops, for instance because
  Resolve will not import a camera file, the program now deletes the
  half-built project it created and opens the one open before.
  Projects it did not create stay untouched.
• The window and the run now refuse an unusable time window with the
  same rule and the same sentence: an Out point before, on or less
  than five seconds after the In point, and a window wholly outside the
  material, naming both marks and the material's length.
• The player's internal "[GUI]" lines now go to the log file only,
  no longer to the Output tab.

Removed

• audio_chain.py, the standalone local sound chain beside the
  program, is gone from the repository. The program never used it.

Fixed

• A recording the recorder split into several files is now written
  down in full in the window, not just its first file: transcript and
  name proposals now reach every block, as the separation already did.

• Two folders now give two speakers even when their files suggest the
  same name: ZOOM0001.WAV from two cards becomes "ZOOM" and "ZOOM
  (2)". Files of one folder still join, and a name typed alike on both
  still joins them.

• When two sound recordings share a file name, the run's log now names
  the second "(2)", as the file list and the preflight already did.
  The plan's summary now counts a track's recordings, not their blocks.

• The "Kind" field now marks as wide shot exactly the camera the run
  cuts to as one. A camera with its own sound and an empty name field
  showed "Wide shot" while the run seated the guessed name on it.

• A camera's own sound set to "no camera of its own" no longer puts
  anybody on that camera, so it can now be the wide shot. The window
  sets it so for every camera sound with voices under it.

• A mark set in the player now lands on the moment marked when the
  player holds a sound file or a camera at another frame rate than the
  reference camera. It was counted at the wrong rate and landed up to a
  few frames late.

• The progress bar now plans the camera-audio stage whenever a
  camera's sound is a track, not only with Multitrack ticked, and no
  longer jumps to nearly full at the start.

• The Auphonic multitrack log now says "Production created" before the
  upload, and "Production running" only once the production has
  started. Before, it said "running" before a single track was up.

• The credit a production needs is now counted in whole minutes,
  rounded up: a 20-second sample needs 1 min, where the line said
  "0 min".

• The progress line now names only the speaker while levelling one:
  "Mixing Guest", no longer "Mixing level_Guest".

Security

• The Auphonic key no longer touches the disc: curl now reads it
  from its input instead of from a temporary file.

Documentation

• The Auphonic chapter now walks through a real run step by step,
  single-track and Multitrack: the messages, where the files land, and
  what it means when auphonic.com refuses a production.
• The public roadmap is gone. What is planned is kept on the owner's
  board, and what each version brings stays in this changelog and on
  the release pages.

Deutsch

Hinzugefügt

• Eine anderswo fertig gemischte Stereo-Mischung kann jetzt den
  Full-Mix ersetzen: „Im Ton“ auf „Fertige Mischung“ stellen oder
  --finished-mix DATEI angeben. Sie wird auf die Zeitachse gelegt und
  kommt unverändert in die Kameradateien und die Resolve-Übergabe; sie
  ist kein Sprecher und geht nicht an auphonic.com.
• Der Auphonic-Kasten zeigt jetzt Tarif und verbleibende Minuten des
  Kontos, sobald das Konto antwortet, auch ohne Preset und mit „ohne
  Auphonic arbeiten“. Die Zeile wird rot, wenn die Aufnahmen mehr
  Minuten brauchen, als übrig sind.
• Die Prüfung vor einem Lauf fragt das Auphonic-Konto jetzt einmal
  und sagt, ob die übrigen Minuten für die Produktion reichen. Im
  kostenlosen Tarif warnt sie bei einer Multitrack-Produktion über
  21 Minuten, dort und noch einmal vor dem Hochladen; der Lauf geht
  trotzdem weiter.
• Jede Veröffentlichung trägt jetzt je System eine Liste der
  Paketversionen, mit denen ihre Testreihe grün lief
  (constraints-macos.txt, -windows.txt, -linux.txt);
  pip3 install -c constraints-macos.txt ... installiert genau diese.

Geändert

• Die Schnittvorschau ist jetzt der Lauf selbst: ein Probelauf
  dessen, was im Fenster steht, angehalten, bevor er schreibt, im
  Hintergrund und ohne Zeile im Protokoll. Vorschau, „Probelauf“ und
  „Start“ schneiden dieselben Einstellungen.
• Ein geänderter Schnittwert rechnet jetzt nur den Schnitt neu. Eine
  geänderte Datei, ein Name, eine Kamera, ein In- oder Out-Punkt lassen
  die Vorschau über dem schon Gemessenen neu laufen.
• „Start“, „Probelauf“ und „Resolve-Projekt anlegen“ warten nicht mehr
  auf eine Vorschau, die noch rechnet: Sie halten deren Lauf an und
  beginnen sofort.
• Nach „Start“ zeigt die Schnittvorschau die Übergabe des fertigen
  Laufs und sagt „aus dem fertigen Lauf“. Ein Lauf mit anderen
  Einstellungen ersetzt sie nicht; ein späterer Probelauf mit denselben
  Einstellungen übernimmt wieder.
• Hat das Fenster seine Niederschrift geschrieben, schneiden Vorschau
  und Lauf nach denselben Wörtern: Wird aus einer Frage in der
  Vorschau ein Reaktionsschnitt, dann auch im Lauf.
• Der „Probelauf“ rechnet jetzt alles bis zum fertigen Schnitt und zur
  Resolve-Übergabe aus und schreibt weiterhin nichts. Der Lauf danach
  nimmt Zeitachse und Sprecher aus dem, was der Probelauf aufbewahrt hat.
• Fenster und Lauf teilen sich jetzt eine Zeitachse: Wer zuerst
  misst, dessen Ergebnis übernimmt der andere, ohne neu zu messen. Eine
  Aufnahme in Blöcken misst auch das Fenster zusammengelegt, wie der Lauf.
• Kameras liegen jetzt gleich, in welcher Reihenfolge sie auch kommen.
  Von gleich langen Kameras ist die Bezugskamera die, die die meisten
  anderen legt, dann eine mit Timecode, dann die erste nach Namen.
• Eine Kamera, die die Bezugskamera nicht hört, wird jetzt über eine
  gelegt, die sie hört, und ihre Zeile im Protokoll sagt, über welche.
  Eine Aufnahme, die nur die letzte Minute mit einer Kamera teilt, wird
  gefunden.
• Liegen Uhr und gemessener Platz einer Kamera mehr als ein Bild
  auseinander, nennt das Protokoll sie jetzt samt Abstand; es gilt die
  Messung. Hat die Bezugskamera keinen Timecode, trägt jede geschriebene
  Kamera den ihres gemessenen Platzes ab 00:00:00.
• Wer über die Spur einer Kamera und über die Stimme einer anderen
  zugeordnet ist, steht jetzt auf der Kamera der Stimme, in
  Schnittliste, Übergabe und Vorschau gleich, und da...

3.0.0b27 -- marks count from the moment every camera runs, and Stop is heard in every stage

Added

• The log now says when a voice sounds like one from another
  recording, with how alike the two are; no name is changed by it.
  Separations stored before are computed once more, the first time.

Changed

• A run from the command line is now levelled to -16 LUFS by default,
  as the window already does. To keep the sound exactly as recorded,
  give --lufs source, the counterpart of "Take from source files".

• From the command line, a preset that masters to anything but
  -16 LUFS now stops the run before the upload, as in the window. Give
  --lufs source or the preset's own number to go on.

• A command-line run that would send to auphonic.com and finds no key
  stored now stops before it starts, not after the time axis. A dry
  run and --without-auphonic need no key.

• Without Multitrack the run now cuts to the cameras chosen under
  "belongs to" and leaves out recordings set to "do not use", as with
  Multitrack. It used to cut everything to the wide shot.

• Without Multitrack, the name typed for a recording is now the name
  the run uses in the log, the track names and the written sound
  files. On the command line --speaker-name FILE NAME does the same.

• Several recordings without any picture are now laid on one time
  axis whether or not Multitrack is ticked, and go to auphonic.com as
  one multitrack production. One whose sound matches nothing is placed
  by its timecode.

• The auphonic.com preset list now offers the kind of preset the run
  will need, tick or no tick. A preset of the wrong kind stops the run
  before the time axis is measured, not after it.

• --auphonic-done now also works without a picture: the processed
  tracks are taken, and nothing is uploaded or charged.

• The command line now refuses a "Sync only" project with more than
  one recording before it starts, as the window already did.

• A voice recognised in another recording may now carry that
  recording's name, and the assignment table proposes it. Any other
  name on two voices still holds "Start", in the window and on the
  command line.

• Without Multitrack, a name typed on two recordings now asks "Names
  used more than once" before they are merged, as with Multitrack.

• With several cameras, no audio recording and Multitrack off,
  "Start" is now greyed out and says why. Each camera would bring its
  own sound, and that is what Multitrack is for.

• When In point or Out point has changed since the last run, "Create
  Resolve project" is now greyed out and says which mark moved and to
  press "Start" again. The command line's Resolve step refuses in the
  same words.

• Where the other voice speaks only once, that turn now holds the
  wide shot only at the nearer edge, and the log says so under "Wide
  shot at the edges".

• Two recordings with the same file name are now told apart by
  "(2)" in the file list, on the Assignment tab and in the preflight,
  as cameras already were.

• The speaker model is now taken from this version's tag, and from
  the main branch only where that tag does not exist. A timeout or a
  server fault stops the fetch with a message naming the tag.

• A run that stops on a fault nobody foresaw now writes the traceback
  into the log and shows one line beginning "Stopped:", in the window
  and on the command line alike.

• A relative In or Out point ("+12:30") saved from the window in an
  earlier version is now read from the moment every camera runs, so
  such a mark can move when the project is opened.

• On Linux, saving the Auphonic key where secret-tool is missing
  now offers to install it with the package manager, in a box in the
  window and as [Y/n] after --store-auphonic-key. Answered no, the
  command to type by hand is named.

Fixed

• Stop now ends a run while the camera files are written, also with
  more cameras than run at once; a camera still waiting started after
  Stop and ran to its end. On Windows, Stop also reaches an ffmpeg
  installed through Chocolatey or Scoop.
• An Out point counted back from the end ("-0:00:20") no longer stops
  a run on Python 3.10 to 3.13; it was read as an unknown switch.
• Stop now ends a run's own speaker separation at once; it used to
  run on until the separation finished, measured 60 s. A single-track
  run waiting for auphonic.com now says so in the bar and in a stop
  report, where it said "Common time axis".
• A recording the recorder split into several files is now separated
  as one recording. The separation heard only the first file, so from
  the second on nobody on that microphone reached the preview or the
  cut.
• Mark In and Mark Out now count from the moment every camera runs,
  as the run does. On material without a clock the run could cut 20
  to 25 seconds after the picture you marked; it now cuts on it.
• A mark set on a camera running at 29.97 frames now lands on its
  frame. It landed three frames late, and preview and run cut three
  frames apart.
• After a run, the Resolve cut tab no longer applies In point and Out
  point a second time to what the run has already cut, and no longer
  says the time window lies outside the material.
• An Out point counted back from the end ("-0:00:30") now means the
  same point in the preview, the player and "Create Resolve project" as
  in the run: back from where the first camera stops.
• Before any run, the preview on the Resolve cut tab now starts and
  ends where the run's Timeline does, so it shows the cut the run will
  build. It began with the earliest recording and could cut its first
  shots differently.
• The proposals for speaker names and "do not use" now count only
  the stretch the run cuts. A voice heard only before every camera ran
  was proposed as a speaker; it is now proposed for "do not use".
• A phone recording with a variable frame rate is now counted at one
  rate everywhere: in the cut, the Timeline, the timecode and on the
  Resolve side. A timecode could land seven frames late.
• A recording that started late now lands in the right place when it
  is placed by phase. One starting 540 s into a 10-minute camera, for
  example, was placed 1048.6 s off.
• The window's speaker measurement now hears every block of a
  recording the recorder split into several files. The preview used
  to show a guest who speaks only in a later part as never speaking.
• The Assignment tab now marks a recording of several blocks red when
  any block does not fit the time axis, as the file list does. It
  looked at the first block only.
• An intro or outro made of several blocks is no longer marked red
  for a later block: the Kind of the first block now holds for all.
• Opening the Resolve cut tab no longer measures the speakers again
  while a stored separation still fits: the same recording, the same
  model and the same separation code.
• After a run without Multitrack, the preview now takes up the run's
  result by itself.
• The disk estimate now counts a recording of several blocks once
  instead of once per block, and reckons the same with and without
  Multitrack.
• The progress bar no longer jumps to 98 % while the camera files are
  still being written without Multitrack.
• Stop pressed while the camera files are written now stops the
  run. It used to go on and write the cut beside camera files a second
  or two long; a file Stop cuts short is now removed.
• Stop now also ends the run while results are downloaded, a preset
  is read, a join is levelled or several files are worked on at once,
  instead of carrying on or ending as "Finished with errors".
• After Stop, the file check and the time axis now work again
  straight away, without pressing "Start" first.
• After Stop, the report now lists only what lies in the output
  folder and names the stage in plain words, such as "Writing the
  camera files", instead of a file name.
• A camera file that cannot be written, for example with the disk
  full, no longer stays in the result folder under its finished name,
  and the cut list leaves that camera out.
• A camera name holding a folder or drive separator is now written as
  a plain name inside the result folder, and the log says so.
• What auphonic.com puts around a returned file, such as the free
  plan's jingle, is now cut away. A 20 s track came back 26.4 s long
  and was kept so; it is now 20.0 s, and the log says what went.
• In Japanese, the YouTube entry of the loudness list now fits its
  field on a Mac.

Security

• onnxruntime, which the speech recognition and the speaker separation
  bring with them, no longer reports to Microsoft. Its telemetry is
  switched off before it loads, and with it goes the thread that could
  make Python quit unexpectedly when a run ended.
• A fetched ffmpeg build is now checked against its release's
  checksums before it is unpacked. Where the list is missing or the
  sum differs, nothing is unpacked; an archive holding a tool twice is
  refused.
• A file name auphonic.com gives a download is now cut to a plain file
  name before it is saved, so nothing lands outside the chosen folder.
  The key now goes to auphonic.com only.
• A list of model files naming a place outside the model folder is now
  refused before anything is fetched or written.

Tests

• Pull requests from forks now run the test suite on the builder.
  Branches of this repository still run it once, on their push.
• No test reaches a DaVinci Resolve open on the machine any more.
  About 23 read-only connections per suite run went to it before.

Documentation

• THIRD-PARTY.md now describes the program as pip installs it, each
  dependency with its licence, and a test holds that list against
  pyproject.toml.
• The chapters on the simple path and on Multitrack now say what the
  tick still decides, and that the camer...

3.0.0b26 -- the window comes apart into pieces and still cuts the same

Added

• A run that goes quiet now says whether it is still working. After
  two minutes without a message the line beside the progress bar says
  "working"; after five minutes in which nothing changes at all it says
  "may be stuck, please check".
• The command line takes --store-auphonic-key. It asks for the
  Auphonic key without showing it, stores it, reads it back and says
  whether it holds.
• On Linux the Auphonic key is now stored in the desktop's keyring
  through secret-tool. Where no keyring answers, nothing is stored and
  the program says so.

Changed

• The camera-cut settings now sit in three groups, "Timing", "Wide
  shot" and "Special cases", that open and close. A closed group shows
  its values in grey, and where the tab would have to scroll, the group
  opened longest ago closes.
• "After a question", "Answer on screen earlier", "Wide shot after"
  and "Wide shot at most" now open before the first run. Once the time
  axis stands, the window writes the transcript in the background, and
  the progress box shows a "Transcript" line.
• Where the transcript is missing, the note under the greyed settings
  now says why: none written yet, speech recognition switched off, or
  the recognition failed. A failure stands in the warning colour, with
  what to do next.
• "Nobody speaks" now holds the picture through a pause of up to one
  second by default instead of going to the wide shot. A project saved
  with "Wide shot" keeps it.
• With "Recognition uncertain" on "No camera change", an uncertain
  passage longer than "Short gap up to" now goes to the wide shot. It
  used to hold the camera however long it ran; without a wide shot it
  still does.
• A clock drift is now only corrected when it is at least three times
  its own uncertainty. A smaller one is left in, and the log line of
  that recording or camera says so and why.
• A camera's drift is now also corrected when the camera runs under
  two minutes or the drift stays under half a frame. Only a drift of
  500 ppm or more is still left in, as rather a measuring error.
• Subtitles in the .srt now hold at most two lines of 42 characters,
  the speaker's name counted in. A long subtitle could run to three
  lines before.
• The preflight now shortens a long file name in the middle with
  "…", so the extension and a "(2)" stay visible; it used to cut the
  end off.
• The command line now refuses a speaker name that stands both on a
  recording and on a separated voice, as the window does. It used to
  merge the two and count every turn twice.
• A launcher or menu entry the program laid under an earlier name of
  its own is now removed when it lays a new one, on macOS and on
  Linux. An entry of any other program stays where it is.

Removed

• AUPHONIC_TOKEN is no longer read. A run from the command line
  now takes the Auphonic key the program has stored, the same one the
  window uses; --store-auphonic-key puts it there.

Fixed

• An empty "In point" and "Out point" now stay empty. The window
  used to fill them from the cameras' clock and hand them to the run;
  now the run takes the whole material, as on the command line.
• The time-axis preview now measures the recordings against the
  longest camera, as the run does. It used to take the longest file,
  and so could show a place the run would not use.
• A multitrack run with only a stored Auphonic key no longer stops as
  if there were no key. Every kind of run now takes the stored key;
  --without-auphonic still leaves it untouched.
• "Stop" now works in every phase of a run, also while a start waits
  for the camera sound and during "Create Resolve project". An upload to
  auphonic.com now stops within two seconds.
• Blocks of one recording whose timecodes lie more than half an hour
  apart are no longer joined into hours of silence. The far block is
  now left out, named in the log, and missing from the plan.
• The log line "Wide shot at the edges" now describes the finished
  cut. An edge that merged into its neighbour is no longer announced,
  and with a single camera the line says the whole cut is the wide shot.
• A track's line "has no picture and is left out" now names only an
  end that really loses something, instead of listing "0:00:00.000 at
  the front" beside the end that does.
• An audio file ffprobe cannot read is now called "not readable" in
  the preflight, with ffprobe's own reason. The run leaves it out and
  goes on, instead of ending in a Python error.
• A recording made of several blocks now keeps its warning colour in
  the file list when one of its blocks does not fit the time axis. A
  later block that fitted used to paint the row back to plain.
• The cameras' colour comparison in the log now names each camera
  after its own file. Two cameras could swap their brightness rows
  before.

Security

• A run from the window no longer leaves a file naming the recordings
  and the speakers in the temp folder. It is now removed however the
  run ends, also when the window is closed during the run.
• On Windows the registry entry holding the Auphonic key is now shut
  to everybody but the user before the key goes in, and the key is
  read back to check it holds.

Tests

• The tests that talk to a running DaVinci Resolve now start only with
  bash resolve.sh --go, and the ones that reach auphonic.com only
  with bash auphonic.sh --online. Without the word they start nothing
  and say which word starts them.

Documentation

• The speech chapter of the manual now describes the three transcript
  files, the .json, the .srt and the .txt, and what each holds.
• The channels chapter now says what to do when a file's channel
  count cannot be read, step by step.

Deutsch

Hinzugefügt

• Ein Lauf, der still wird, sagt jetzt, ob er noch arbeitet. Nach
  zwei Minuten ohne Meldung steht neben dem Fortschrittsbalken
  „arbeitet“; ändert sich fünf Minuten lang gar nichts, heißt es dort
  „hängt womöglich, bitte prüfen“.
• Die Kommandozeile kennt --store-auphonic-key. Der Schalter fragt
  den Auphonic-Schlüssel verdeckt ab, legt ihn ab, liest ihn zurück und
  sagt, ob er hält.
• Unter Linux kommt der Auphonic-Schlüssel jetzt in den Schlüsselbund
  des Desktops, über secret-tool. Antwortet keiner, wird nichts
  abgelegt, und das Programm sagt es.

Geändert

• Die Einstellungen für den Kameraschnitt stehen jetzt in drei
  Gruppen, „Timing“, „Weitwinkel“ und „Sonderfälle“, die sich auf- und
  zuklappen lassen. Eine zugeklappte Gruppe zeigt ihre Werte grau;
  müsste der Reiter rollen, klappt die am längsten offene zu.
• „Nach einer Frage“, „Antwort früher im Bild“, „Weitwinkel nach“ und
  „Weitwinkel höchstens“ sind jetzt schon vor dem ersten Lauf offen.
  Steht die Zeitachse, schreibt das Fenster die Niederschrift im
  Hintergrund, und der Vorarbeitskasten zeigt eine Zeile „Niederschrift“.
• Fehlt die Niederschrift, sagt der Hinweis unter den ausgegrauten
  Einstellungen jetzt, warum: noch keine geschrieben, Spracherkennung
  abgeschaltet oder Erkennung gescheitert. Das Scheitern steht in der
  Warnfarbe, zusammen mit dem nächsten Schritt.
• „Niemand redet“ hält das Bild jetzt von Haus aus über eine Pause von
  bis zu einer Sekunde, statt auf den Weitwinkel zu gehen. Ein mit
  „Weitwinkel“ gespeichertes Projekt behält das.
• Steht „Erkennung unsicher“ auf „Kein Kamerawechsel“, geht eine
  unsichere Stelle, die länger als „Kurze Lücke bis“ dauert, jetzt auf
  den Weitwinkel. Vorher blieb die Kamera beliebig lange stehen; ohne
  Weitwinkel bleibt sie das weiterhin.
• Ein Uhrengang wird jetzt nur noch herausgerechnet, wenn er
  mindestens dreimal so groß ist wie seine eigene Unsicherheit. Ein
  kleinerer bleibt stehen, und die Protokollzeile der Aufnahme oder
  Kamera sagt das und warum.
• Der Gang einer Kamera wird jetzt auch herausgerechnet, wenn sie
  kürzer als zwei Minuten läuft oder er unter einem halben Bild bleibt.
  Nur ab 500 ppm bleibt er stehen – das ist eher ein Messfehler.
• Untertitel in der .srt haben jetzt höchstens zwei Zeilen zu 42
  Zeichen, der Name des Sprechers mitgezählt. Vorher konnte ein langer
  Untertitel auf drei Zeilen wachsen.
• Der Vorflug kürzt einen langen Dateinamen jetzt in der Mitte mit
  „…“, damit die Endung und ein „(2)“ sichtbar bleiben; vorher fiel das
  Ende weg.
• Die Kommandozeile lehnt jetzt einen Sprechernamen ab, der zugleich
  an einer Aufnahme und an einer getrennten Stimme steht, wie das
  Fenster. Vorher legte sie beide zusammen und zählte jeden Beitrag
  doppelt.
• Einen Starter oder Menüeintrag, den das Programm unter einem
  früheren eigenen Namen angelegt hat, entfernt es jetzt, sobald es
  einen neuen anlegt, unter macOS und Linux. Einträge anderer
  Programme bleiben unberührt.

Entfernt

• AUPHONIC_TOKEN wird nicht mehr gelesen. Ein Lauf von der
  Kommandozeile nimmt jetzt den Auphonic-Schlüssel, den das Programm
  abgelegt hat, denselben wie das Fenster; dorthin kommt er mit
  --store-auphonic-key.

Behoben

• Ein leerer „In-Punkt“ und „Out-Punkt“ bleiben jetzt leer. Das
  Fenster füllte sie vorher aus der Uhr der Kameras und gab sie dem Lauf
  mit; jetzt nimmt der Lauf das ganze Material, wie auf der Kommandozeile.
• Die Vorschau der Zeitachse misst die Aufnahmen jetzt an der längsten
  Kamera, wie der Lauf. Vorher nahm sie die längste Datei und konnte
  so einen Platz zeigen, den der Lauf nicht nimmt.
• Ein Multitrack-Lauf, der nur einen abgelegten Auphonic-Schlüssel
  hat, bricht nicht mehr ab, als gäbe es keinen. Jetzt nimmt jede Art
  von Lauf den abgelegten Schlüssel; --without-auphonic lässt ihn
  weiterhin unberührt.
• „Abbrechen“ wirkt jetzt in jeder Phase eines Laufs, auch während
  ein Start auf den Kameraton wartet und bei „Resolve-Projekt anlegen“.
  Ein Ho...

3.0.0b25 -- every recording says what is in its sound, and a copied project saves into itself

Added

• Every recording in the file list now has a field "In the sound":
  "Speech" or "Mixed". Set a recording with music or a mix under the
  voices to Mixed, and the program may place it by the phase of its
  sound.
• The command line takes --sound speech|mixed for every recording
  and --sound-of FILE WHAT for one, the same answer the "In the
  sound" field gives in the window.

Changed

• A recording on "Speech" that shares no sound with the cameras is now
  refused instead of being placed by the phase of its sound, which on
  speech put a recording from elsewhere far off the mark.
• A recording that no way of measuring can place now stands at its
  timecode, in the run and in the preview alike, instead of where the
  failed guess of its sound would put it.
• After auphonic.com, who speaks when is now worked out by voice on
  the raw microphones, as it is without the service. On four real
  productions that found the right speaker 99 % of the time, against
  about 80 % from the processed tracks.
• "Wide shot for greeting at the start and farewell at the end" now
  shortens an edge of a short recording instead of dropping it. Each
  edge is cut to at most a third of the length, and the log says so.
• Each wide-shot edge now also ends at "Wide shot at the latest",
  120 seconds by default. In a 30-minute recording with a long greeting,
  each end used to stay on the wide shot for ten minutes.
• "Answer on screen earlier" now also brings forward an answer that
  begins a little off the camera change, up to the reaction gap of
  three seconds; before, only an answer starting exactly there counted.
• The log, the plan and the preflight now name the second of two
  camera files of one name "C0003.MP4 (2)", as the window does.
• A camera or voice name typed into the table now stays exactly as
  typed, even one that looks like the table's own proposal, such as
  "Wide 2".
• The window can now be made as narrow as its contents allow, about
  560 pixels in English, instead of stopping at a fixed 1000 × 520; in
  a wide font it no longer squeezes a choice box below its text.
• Transcripts and speaker separations now stay in the cache for 30
  days after they were last used, not after they were made, so a
  recording used every week is not listened to again.

Removed

• The command-line switch --auphonic-api-key is gone. A run from
  the command line takes the key from AUPHONIC_TOKEN, so the key no
  longer stands in the process list or the shell history.

Fixed

• Saving a project opened from a copy of its file now writes into that
  copy. Opening and saving "Beta copy.json" used to overwrite the
  original "Beta.json".
• A production without a file of its own no longer writes over another
  project's file of the same name. It is now saved as "Name (2)", and
  the window says so.
• Renaming a production to another project's name no longer
  overwrites that project's file. The window now says which project
  lies there, and saving goes on into the current file.
• Renaming a production now moves its project file only once the name
  is settled: on leaving the field, on Enter, on "Save project" or at
  the start of a run, instead of on every keystroke.
• Closing a project no longer lets a time axis still being measured
  land in the next production's file; the late answer is now dropped.
• A project saved before its type was chosen now reopens with "Project
  type" still open, and "Start" and "Dry run" wait for the answer
  instead of treating it as "Cut by speaker".
• A project file that names its spoken language in another spelling,
  such as "de" or "fr", now keeps it when the project is opened and
  saved; it used to come back with no language at all.
• --speech-language de on the command line now tags the track
  ger, as the window does; the same holds for deu and for
  --speech-language-camera. ffmpeg used to drop the other spelling and
  leave the track untagged.
• A dry run from the window no longer rewrites the project file in the
  output folder; the folder now stays exactly as it was.
• "Start again now" after an update or the ffmpeg install now reopens
  the window. It used to repeat the last run from the command line and
  upload to auphonic.com without asking.
• After a run, or after the output folder is reset, the window now
  looks for finished Auphonic tracks again, so the next run uses them
  instead of uploading and spending credit a second time.
• "Overwrite files" no longer asks about files an earlier run of this
  production wrote, and it now names the camera files the run really
  writes.
• A camera much shorter than the others is now found wherever it
  begins. It used to be looked for only near the start of the longest
  camera, and refused wherever it began later.
• The preview now refuses a short camera from another room, as the run
  does, instead of placing it where the run then left it out.
• A recording block that fits weakly on too few points is no longer
  laid out half a minute wrong; it now stands at its clock instead.
• A run with --auphonic-done now refuses a recording that lies in
  that same folder and names the file, instead of running through with
  the speakers read off the returned tracks.
• A recording joined from several microphones now keeps its speaker
  separation after 30 days, instead of being separated again when the
  joined file is built afresh.
• Without "Multitrack", the handover and the lists are now named after
  the production, not after the folder the material lies in.
• A file set to "ignore this video" is no longer counted under the
  file list as one that does not fit.
• A file that stops being marked as not fitting shows its folder again
  in the file list instead of the old note.
• On Windows the program now ends cleanly when its window closes;
  with Python 3.10 it used to crash on the way out.
• In Arabic, Persian, Hebrew and Urdu every line of the log now reads
  from the right, also a line that begins with a file name.
• In Arabic, Persian, Hebrew and Urdu a signed number now keeps its
  sign in front: the log and the preflight read "-16 LUFS" where they
  read "16- LUFS".

Security

• The window now hands the Auphonic key to its run inside the
  program, so the key no longer stands in the process list while the
  run lasts.

Documentation

• The manual now describes the "In the sound" field and says when
  the program places a recording by the phase of its sound.
• The manual shows a command-line run as text instead of a terminal
  picture.

Deutsch

Hinzugefügt

• Jede Aufnahme in der Dateiliste hat jetzt ein Feld „Im Ton“:
  „Sprache“ oder „Gemischt“. Liegt Musik oder eine Mischung unter den
  Stimmen, stellt man sie auf Gemischt, und das Programm darf sie über
  die Phase ihres Tons platzieren.
• Die Kommandozeile kennt --sound speech|mixed für alle Aufnahmen
  und --sound-of DATEI WAS für eine, dieselbe Antwort, die im
  Fenster das Feld „Im Ton“ gibt.

Geändert

• Eine Aufnahme auf „Sprache“, die keinen Ton mit den Kameras teilt,
  wird jetzt abgelehnt, statt über die Phase ihres Tons platziert zu
  werden – bei Sprache landete so eine fremde Aufnahme weit daneben.
• Eine Aufnahme, die kein Messweg platzieren kann, steht jetzt an
  ihrem Timecode, im Lauf wie in der Vorschau, statt dort, wohin die
  gescheiterte Schätzung ihres Tons sie gelegt hätte.
• Nach auphonic.com wird jetzt an den rohen Mikrofonen über die Stimme
  bestimmt, wer wann spricht, genau wie ohne den Dienst. In vier
  echten Produktionen traf das zu 99 % den richtigen Sprecher, über die
  bearbeiteten Spuren nur zu etwa 80 %.
• „Weitwinkel für Begrüßung am Anfang und Verabschiedung am Ende“
  kürzt bei einer kurzen Aufnahme jetzt den Rand, statt ihn zu
  streichen. Jeder Rand wird auf höchstens ein Drittel der Länge
  gekürzt, und das Protokoll sagt es.
• Jeder Weitwinkel-Rand endet jetzt auch bei „Weitwinkel
  spätestens“, in der Grundeinstellung nach 120 Sekunden. In einer
  30-minütigen Aufnahme mit langer Begrüßung blieb vorher jedes Ende
  zehn Minuten auf dem Weitwinkel.
• „Antwort früher im Bild“ zieht jetzt auch eine Antwort vor, die ein
  Stück neben dem Kamerawechsel beginnt, bis zur Reaktionsspanne von
  drei Sekunden; vorher zählte nur eine, die genau dort anfing.
• Protokoll, Plan und Vorflug nennen die zweite von zwei gleichnamigen
  Kameradateien jetzt „C0003.MP4 (2)“, so wie das Fenster.
• Ein in die Tabelle getippter Kamera- oder Stimmenname bleibt jetzt
  genau so stehen, auch wenn er wie ein Vorschlag der Tabelle aussieht,
  etwa „Wide 2“.
• Das Fenster lässt sich jetzt so schmal ziehen, wie sein Inhalt
  erlaubt, auf Englisch etwa 560 Pixel, statt bei festen 1000 × 520 zu
  enden; bei breiter Schrift drückt es keine Auswahl mehr unter ihren
  Text zusammen.
• Niederschriften und Sprechertrennungen bleiben jetzt 30 Tage nach
  ihrer letzten Nutzung im Zwischenspeicher, nicht nach ihrer
  Entstehung – eine Aufnahme, die jede Woche gebraucht wird, wird nicht
  noch einmal abgehört.

Entfernt

• Der Kommandozeilenschalter --auphonic-api-key ist weg. Ein Lauf
  von der Kommandozeile nimmt den Schlüssel aus AUPHONIC_TOKEN; so
  steht er weder in der Prozessliste noch im Verlauf der Shell.

Behoben

• Ein Projekt, das aus einer Kopie seiner Datei geöffnet wurde,
  speichert jetzt in diese Kopie. Wer „Beta Kopie.json“ öffnete und
  speicherte, überschrieb vorher das Original „Beta.json“.
• Eine Produktion ohne eigene Datei überschreibt die gleichnamige
  Datei eines anderen Projekts nicht mehr. Sie wird jetzt als
  „Name (2)“ gespeichert, und das Fenster sagt es.
• Wer eine Produktion auf den Namen eines anderen Projekts umbenennt,
  überschreibt dessen Datei nicht mehr. Das F...

3.0.0b24 -- 96 kHz blocks join whole, and the field says it is the language of the sound

Changed

• The field for the language spoken in the recording is now called
  "Language of the sound", so it can no longer be taken for the
  language of the window.
• Three more sentences now name the second of two camera files of one
  name "C0003.MP4 (2)": the note on who holds Intro or Outro, the hint
  beside the greyed In and Out buttons, and the refusal of two files set
  to Intro.

Fixed

• A recording split into timecoded blocks at 44.1 or 96 kHz is now
  joined whole. Two 96 kHz blocks used to come out a quarter short,
  with a gap reported that was not there.

Tests

• A real run's Resolve handover is now checked on material like a real
  production: two frame rates, one camera in 10 bit, cameras nobody
  speaks on.

Documentation

• The manual now shows the "Output" tab as a dry run leaves it.

Deutsch

Geändert

• Das Feld für die Sprache, die in der Aufnahme gesprochen wird, heißt
  jetzt „Sprache im Ton“, damit es niemand mehr für die Sprache des
  Fensters hält.
• Drei weitere Sätze nennen die zweite von zwei gleichnamigen
  Kameradateien jetzt „C0003.MP4 (2)“: der Hinweis, wer Intro oder
  Outro belegt, der Hinweis neben den ausgegrauten In- und Out-Knöpfen und
  die Ablehnung zweier Dateien, die beide auf Intro stehen.

Behoben

• Eine Aufnahme in Timecode-Blöcken mit 44,1 oder 96 kHz wird jetzt ganz
  zusammengefügt. Zwei 96-kHz-Blöcke kamen vorher ein Viertel zu kurz
  heraus, mit einer gemeldeten Lücke, die es nicht gab.

Tests

• Die Resolve-Übergabe eines echten Laufs wird jetzt an Material wie
  aus einer echten Produktion geprüft: zwei Bildraten, eine Kamera in
  10 Bit, Kameras, auf denen niemand spricht.

Dokumentation

• Das Handbuch zeigt jetzt den Reiter „Ausgabe“, wie ein Probelauf ihn
  hinterlässt.

3.0.0b23 -- every tab fits 1280 px in every language, Tamil on Linux included

Changed

• The row with the production's name, language and project type now
  continues on a second line where the window is too narrow, instead
  of pushing the "Files & production" tab past its edge.
• On the "Assignment & time window" and "Resolve cut" tabs the preview
  now goes below the settings when the two columns do not fit side by
  side. In Tamil all three tabs now fit a 1280 px screen on Linux too.
• The note on a second wide shot now names the camera the cut takes as
  the window does, "C0003.MP4 (2)" for the second of two files of one
  name.

Fixed

• Bit depths now read in the chosen language in the file list's
  colour row, the preflight and the HDR check; German says "8 Bit"
  instead of "8 bit".
• A floating-point recording is now described as "32 bit float"
  instead of "32f bit".
• The render settings in the Resolve log now write the frame rate with
  the language's decimal mark.

Deutsch

Geändert

• Die Zeile mit Name, Sprache und Projekttyp der Produktion bricht
  jetzt in eine zweite Zeile um, wo das Fenster zu schmal ist, statt
  den Reiter „Dateien & Produktion“ über seinen Rand hinauszuschieben.
• In den Reitern „Zuordnung & Zeitfenster“ und „Resolve-Schnitt“ rückt
  die Vorschau jetzt unter die Einstellungen, wenn die zwei Spalten
  nicht nebeneinander passen. Auf Tamil passen alle drei Reiter jetzt auch
  unter Linux auf 1280 px.
• Der Hinweis beim zweiten Weitwinkel nennt die Kamera, die der Schnitt
  nimmt, jetzt so wie das Fenster: „C0003.MP4 (2)“ für die zweite von
  zwei gleichnamigen Dateien.

Behoben

• Bittiefen stehen jetzt in der gewählten Sprache in der Farbzeile der
  Dateiliste, in der Vorprüfung und in der HDR-Prüfung; Deutsch sagt
  „8 Bit“ statt „8 bit“.
• Eine Gleitkomma-Aufnahme heißt jetzt „32 Bit Gleitkomma“ statt
  „32f bit“.
• Die Render-Einstellungen im Resolve-Protokoll schreiben die Bildrate
  jetzt mit dem Dezimalzeichen der Sprache.

3.0.0b22 -- two cameras of one name are told apart, and the limiter catches the first click

Changed

• Two camera files with the same name are now told apart by their
  path. The window lists the second as "C0003.MP4 (2)", with the path
  as a tooltip; an older project gives both rows to the first.
• "Create Resolve project" is offered again after reopening a project
  whose run left a camera out. A camera added after that run keeps
  the button grey.
• Under "Sync only" the greyed preview and the wide-shot note on the
  Resolve tab now stand empty instead of speaking of speakers and a
  wide shot.
• The preflight now notes two cameras that are one recording added
  twice. It is a hint and stops nothing.
• A separation file named with --speakers-from is no longer used
  when its recording has changed or it came from another model; the
  run warns and goes on without it.
• The cache now drops transcripts and speaker separations written
  more than 30 days ago, and old builds of the speech recogniser.

Fixed

• The limiter now holds a peak at the very start of a recording under
  -1 dBTP. Such a peak used to pass at full scale.
• On a 1280 px screen the "Resolve cut" tab no longer scrolls
  sideways; the explanations beside the settings now wrap.
• A long output folder is now shortened in the middle instead of
  pushing the "Files & production" tab past the window; the tooltip holds it whole.
• The two buttons in the empty window now grow with their text
  instead of cutting a longer translation off.
• With both tools missing the message now reads "ffmpeg and ffprobe
  are missing." instead of "ffmpeg, ffprobe is missing."
• The first lines of a run now stand in its own log, and an empty
  log no longer overwrites the kept one.
• The camera's colour boxes are now taken back if ffprobe can no
  longer open the file after they were copied in.
• The Resolve build no longer asks for audio tracks for a camera
  whose video track was refused.
• The preview now offers the outside player also where the system
  only reports a file as unopenable.

Tests

• Eight paths no test entered are now tested, among them tracks
  that come back from Auphonic, the limiter and the refusals of a run.

Deutsch

Geändert

• Zwei Kameradateien mit gleichem Namen werden jetzt über ihren Pfad
  unterschieden. Die zweite heißt im Fenster „C0003.MP4 (2)“, der Pfad
  steht im Tooltip; ein älteres Projekt gibt beide Zeilen der ersten.
• „Resolve-Projekt anlegen“ steht nach dem Wiederöffnen eines
  Projekts wieder bereit, dessen Lauf eine Kamera ausgelassen hat. Eine
  danach hinzugefügte Kamera lässt den Knopf grau.
• Bei „Nur synchronisieren“ bleiben die ausgegraute Vorschau und der
  Weitwinkel-Hinweis im Reiter Resolve jetzt leer, statt von Sprechern
  und einem Weitwinkel zu reden.
• Die Vorprüfung meldet jetzt zwei Kameras, die eine doppelt
  hinzugefügte Aufnahme sind. Es ist ein Hinweis und hält nichts auf.
• Eine Sprechertrennung aus --speakers-from wird nicht mehr
  verwendet, wenn sich ihre Aufnahme geändert hat oder sie von einem
  anderen Modell stammt; der Lauf warnt und macht ohne sie weiter.
• Der Zwischenspeicher löscht jetzt Transkripte und Sprechertrennungen,
  die vor mehr als 30 Tagen geschrieben wurden, und alte Bauten der
  Spracherkennung.

Behoben

• Der Limiter hält eine Spitze ganz am Anfang einer Aufnahme jetzt
  unter -1 dBTP. Sie ging vorher ungebremst durch.
• Auf einem 1280 px breiten Bildschirm scrollt der Reiter
  „Resolve-Schnitt“ nicht mehr seitwärts; die Erklärungen brechen um.
• Ein langer Ausgabeordner wird jetzt in der Mitte gekürzt, statt
  den Reiter „Dateien & Produktion“ über das Fenster hinauszuschieben; der Tooltip
  zeigt ihn ganz.
• Die zwei Knöpfe im leeren Fenster wachsen jetzt mit ihrem Text,
  statt eine längere Übersetzung abzuschneiden.
• Fehlen beide Werkzeuge, heißt die Meldung jetzt „ffmpeg und ffprobe
  fehlen.“ statt „ffmpeg, ffprobe fehlt.“
• Die ersten Zeilen eines Laufs stehen jetzt in seinem eigenen
  Protokoll, und ein leeres Protokoll überschreibt das aufbewahrte nicht
  mehr.
• Die Farbangaben der Kamera werden jetzt zurückgenommen, wenn
  ffprobe die Datei nach dem Einkopieren nicht mehr öffnet.
• Der Resolve-Bau fordert jetzt keine Tonspuren mehr für eine Kamera
  an, deren Videospur abgelehnt wurde.
• Die Vorschau bietet den externen Abspieler jetzt auch an, wenn das
  System eine Datei nur als nicht zu öffnen meldet.

Tests

• Acht Wege, die kein Test betrat, werden jetzt geprüft, darunter
  Spuren, die von Auphonic zurückkommen, der Limiter und die Ablehnungen
  eines Laufs.

3.0.0b21 -- the tests move into their pieces' folders, and a second project leaves the first alone

Changed

• Two camera files with the same name no longer lock Start. The
  window names the second one " 2", as the command line does, and
  each keeps its own picture.
• Under "Sync only" no camera is made the wide shot because nobody
  speaks on it. Only a camera marked as the wide shot is one, in the
  window and in the handover.
• A camera's own sound now always stands where its camera stands,
  however the camera was placed; it is no longer measured a second time.
• A camera Resolve would not insert now ends the build with errors,
  like a track refused twice: the log says it again at the end, the
  window ends on "Finished with errors", and --resolve ends in an
  error code.
• Before the Auphonic mix exists, the wide shot in the preview plays
  the one recording that carries every voice, not the camera's own
  sound.

Fixed

• Opening a project no longer damages another project's file. A
  second project opened in the same window, or an unnamed production's
  file beside the one opened, used to lose its file or its production.
• The line under the file list no longer says "nothing to fault"
  while a file does not fit the others. It now counts those files.
• "to In point" and "to Out point" now say when a timecode lies
  outside every video file, also with a timecoded camera in the
  player, instead of jumping silently to an end.
• A recorder file at 44.1 kHz now starts at the timecode it carries.
  Its start used to be read about 8 % too early.
• Where no clock drift was measured, the log now says "clock drift not
  measured" instead of a line of noughts that read like a measurement.
• The title bar now follows a renamed production.
• The file details now show their units and headings in the chosen
  language -- sizes, bit depth, sample rate, frame rate, mono and
  stereo.

Tests

• The tests now stand in folders by program piece, tests/cut/,
  tests/ui/ and so on; the four that need a running DaVinci Resolve
  stand in tests/resolve/live/.

Deutsch

Geändert

• Zwei Kameradateien mit gleichem Namen sperren „Start“ nicht mehr.
  Das Fenster nennt die zweite „ 2“, wie die Kommandozeile, und
  jede behält ihr eigenes Bild.
• Bei „Nur synchronisieren“ wird keine Kamera zum Weitwinkel, nur weil
  niemand auf ihr spricht. Weitwinkel ist nur, was als solcher markiert
  ist, im Fenster wie in der Übergabe.
• Der eigene Ton einer Kamera steht jetzt immer dort, wo die Kamera
  steht, gleich wie sie eingeordnet wurde; er wird nicht mehr ein
  zweites Mal gemessen.
• Eine Kamera, die Resolve nicht einfügt, beendet den Bau jetzt mit
  Fehlern, wie eine zweimal abgelehnte Spur: Das Protokoll wiederholt
  es am Ende, das Fenster sagt „Mit Fehlern beendet“, --resolve
  endet mit Fehlercode.
• Solange es den Auphonic-Mix noch nicht gibt, spielt der Weitwinkel
  in der Vorschau die eine Aufnahme, die alle Stimmen trägt, nicht den
  Ton der Kamera.

Behoben

• Ein Projekt zu öffnen, beschädigt jetzt keine andere Projektdatei.
  Ein zweites Projekt im selben Fenster, oder die Datei einer
  unbenannten Produktion neben dem geöffneten, verlor bisher seine Datei
  oder seine Produktion.
• Die Zeile unter der Dateiliste sagt nicht mehr „nichts zu bemängeln“,
  während eine Datei nicht zu den anderen passt. Sie zählt diese
  Dateien jetzt.
• „zu In-Punkt“ und „zu Out-Punkt“ sagen jetzt, wenn ein Timecode in
  keiner Videodatei liegt, auch mit einer Kamera mit Timecode im
  Abspieler, statt still an ein Ende zu springen.
• Eine Rekorderdatei mit 44,1 kHz beginnt jetzt bei dem Timecode, den
  sie trägt. Ihr Anfang wurde etwa 8 % zu früh gelesen.
• Wo kein Uhrengang gemessen wurde, sagt das Protokoll jetzt „Uhrengang
  nicht gemessen“ statt einer Zeile aus Nullen, die wie eine Messung
  aussah.
• Die Titelleiste folgt jetzt einer umbenannten Produktion.
• Die Dateiangaben zeigen ihre Einheiten und Überschriften jetzt in
  der gewählten Sprache -- Größe, Bittiefe, Abtastrate, Bildrate, Mono
  und Stereo.

Tests

• Die Tests stehen jetzt in Ordnern je Programmteil, tests/cut/,
  tests/ui/ und so weiter; die vier, die ein laufendes DaVinci Resolve
  brauchen, in tests/resolve/live/.

3.0.0b20 -- a camera that cannot hear goes by its clock, in forty-eight languages

Added

• The window and the log now speak 48 languages. Nineteen are new,
  from Bosnian to Urdu, under "Language" and as --lang. In Persian,
  Hebrew and Urdu the window reads from right to left, as in Arabic.

Changed

• A camera its sound cannot place now stands where its timecode says,
  set against a camera the sound did place. It used to land at a guess
  from the sound, in one measured case three seconds off. Without one it
  stays out, and the program says why.
• Several speakers on one camera now stand in alphabetical order
  everywhere, capitals and accents aside: in the file name, the
  Resolve track, "gets audio from" and inside the file. An older project
  shows and writes them in the new order.
• Under "Sync only" a camera is proposed its own file name as its
  "new file name", never a name made from speakers.
• A locked field now says why, in grey, inside itself -- "no
  speaker" in the Kind of a wide shot, "no audio track" in a locked
  "Camera audio", "not used" in the name of an unused recording. It used
  to stand in a tooltip or another column.
• When Resolve refuses a track, the program asks a second time.
  Refused again, the log names the track and what the Timeline lacks,
  the rest is built, and "Create Resolve project" ends on "Finished with
  errors" instead of "Done".
• "Create Resolve project" follows the file list. Taking a camera
  out greys it when the last handover no longer names exactly these
  cameras; putting it back makes it usable again, where that handover
  lies in the output folder or beside the project.

Fixed

• A camera with no sound track no longer drops out without a word.
  It is named and placed by its timecode where it has one -- in the
  preview and now in the run too, which used to leave it out of the
  Resolve handover.
• The handover no longer says "Nothing was found in the sound" of a
  camera whose sound matched, only too weakly. It now says which of
  the two it was.
• The colour comparison no longer warns about a 10- or 12-bit camera
  beside an 8-bit one. Every camera is now read on the 8-bit scale.
• A file left out while no Intro is set is no longer told that the
  Intro is taken. It now reads that Intro or Outro is one click away.
• When the key could not be saved and no preset fits the mode, both
  sentences now stay in the settings. The note about the presets used
  to replace the one about the key.
• The line under the preview picture now follows "hear assigned
  audio" at once, also while paused. It kept its old text until the
  picture moved.
• The drop-downs of the camera cut now share one width, that of
  their longest entry. "Answering speaker" was cut off.
• "In point", "Out point", "VIDEO", "Registry", "Status" and
  "Presets:" are now translated. They stood in English in every
  language.

Deutsch

Hinzugefügt

• Fenster und Protokoll sprechen jetzt 48 Sprachen. Neunzehn sind
  neu, von Bosnisch bis Urdu, unter „Sprache“ und als --lang. Auf
  Persisch, Hebräisch und Urdu läuft das Fenster von rechts nach links,
  wie auf Arabisch.

Geändert

• Eine Kamera, die ihr Ton nicht einordnen kann, steht jetzt dort, wo
  ihr Timecode sagt, gemessen an einer vom Ton eingeordneten Kamera.
  Bisher lag sie an einer Schätzung, einmal drei Sekunden daneben. Ohne
  eine solche bleibt sie draußen, und das Programm sagt warum.
• Mehrere Sprecher auf einer Kamera stehen jetzt überall
  alphabetisch, ohne Blick auf Groß- und Kleinschreibung oder Akzente:
  im Dateinamen, in der Resolve-Spur, in „bekommt Audio von“ und in der
  Datei. Ein älteres Projekt zeigt und schreibt sie in der neuen
  Reihenfolge.
• Bei „Nur synchronisieren“ bekommt eine Kamera ihren eigenen
  Dateinamen als „neue Datei heißt“ vorgeschlagen, nie einen Namen aus
  Sprechern.
• Ein gesperrtes Feld sagt jetzt grau in sich selbst, warum -- „kein
  Sprecher“ im „Typ“ eines Weitwinkels, „keine Tonspur“ in einem
  gesperrten „Kameraton“, „ungenutzt“ im Namen einer ungenutzten
  Aufnahme. Bisher stand das im Tooltip oder in einer anderen Spalte.
• Lehnt Resolve eine Spur ab, fragt das Programm ein zweites Mal.
  Lehnt Resolve wieder ab, nennt das Protokoll die Spur und was der
  Timeline fehlt, der Rest wird gebaut, und „Resolve-Projekt anlegen“
  endet mit „Mit Fehlern beendet“ statt „Fertig“.
• „Resolve-Projekt anlegen“ folgt der Dateiliste. Nimmt man eine
  Kamera heraus, wird der Knopf grau, sobald die letzte Übergabe nicht
  mehr genau diese Kameras nennt; legt man sie zurück, ist er wieder
  benutzbar, wo diese Übergabe im Ausgabeordner oder neben der
  Projektdatei liegt.

Behoben

• Eine Kamera ohne Tonspur fällt nicht mehr wortlos heraus. Sie wird
  genannt und nach ihrem Timecode gelegt, wo sie einen hat -- in der
  Vorschau und jetzt auch im Lauf, der sie bisher nicht an Resolve
  übergab.
• Die Übergabe sagt nicht mehr „im Ton nichts gefunden“ über eine
  Kamera, deren Ton passte, nur zu schwach. Sie sagt jetzt, welcher
  der beiden Fälle es war.
• Der Farbvergleich warnt nicht mehr vor einer 10- oder 12-Bit-Kamera
  neben einer 8-Bit-Kamera. Jede Kamera wird jetzt auf der
  8-Bit-Skala gelesen.
• Eine Datei, die draußen bleibt, während kein Intro gesetzt ist,
  hört nicht mehr, das Intro sei vergeben. Sie liest jetzt, dass Intro
  oder Outro einen Klick entfernt sind.
• Ließ sich der Schlüssel nicht speichern und passt kein Preset zum
  Modus, bleiben in den Einstellungen jetzt beide Sätze stehen. Der
  Hinweis zu den Presets ersetzte bisher den zum Schlüssel.
• Die Zeile unter dem Vorschaubild folgt „zugeordneten Ton hören“
  jetzt sofort, auch bei Pause. Sie behielt ihren alten Text, bis sich das
  Bild bewegte.
• Die Auswahllisten des Kameraschnitts sind jetzt gleich breit, so
  breit wie ihr längster Eintrag. „Antwortender“ war abgeschnitten.
• Die Wörter für In- und Out-Punkt, die Überschrift „VIDEO“,
  „Registry“, „Status“ und „Presets:“ werden jetzt übersetzt. Sie
  standen in jeder Sprache englisch.