Help
The questions that come up. If yours is not here, write to us : useful answers end up on this page.
By service
Each service has its own questions. Cross-cutting topics (licences, files, teamwork) follow further down.
Voice / M&E separation
Studios handed a mix with no separated elements · the service page
A real stem, not a leftover
- Is the result deliverable as-is?
- Through the local route, yes: it is a stem. A second route, via an online service, only isolates the voice and rebuilds the M&E by subtraction, fine for a mock-up or an edit, not for broadcast delivery. Axeho names the files accordingly, and the distinction survives the trip into a delivery folder.
- Do my rushes leave for the internet?
- Not through the local route: nothing leaves the workstation. That is what makes it usable on embargoed content.
Filed where the render will find it
- Can I choose another folder?
- Yes, and your choice wins. The default location is only a default: it exists so the two families of tracks meet, not to impose a folder tree on you.
Batch processing
- How long for one episode?
- It depends entirely on your machine: your processor does the work, not a server. Expect several minutes per episode on an ordinary workstation, and run the season overnight.
Rythmo band
Adapters, dubbing directors, actors working from home studios · the service page
Frame-accurate alignment
- How do I move a block by a single frame?
- Ctrl and the left/right arrows move the end of the block. Ctrl+Shift moves the start, Ctrl+Alt moves the whole block without changing its length. The shortcut works even when the cursor is inside the block’s text.
- My timecode starts at 10:00:00:00. Is that normal?
- Yes, that is the usual dubbing reference, and it is adjustable in the band toolbar. An episode starting at zero therefore displays as 10:00:00:00.
Actor mode
- Does the actor see a band re-timed to their lines only?
- No, and that is deliberate. A compressed band would no longer match the picture and the recording would land off. Other characters’ lines are hidden; the duration and timecodes stay those of the episode.
- Can I export one character’s text alone?
- Yes: the role sheet gives their lines in film order, timecoded, with the word count, useful when an actor is paid by the line.
- In practice, how does an actor record from home?
- They open the episode, pick their role, and click **🎬 Enregistrer la bande** (record the band): the video starts, its own sound is muted automatically so it does not bleed into the take, and the microphone records continuously while the band scrolls. One file comes out at the end, with every line placed at its timecode — no second piece of software, and nothing to realign afterwards.
- Does it produce one file per line?
- No. **🎬 Enregistrer la bande** produces **one single file** for the whole session, every line placed at its timecode in the episode — not a separate per-line take to stitch back together in the edit.
- Can I pause partway through recording?
- Yes: **⏸ Pause** stops right where you are, without sending anything. During the pause you can listen to a line already recorded or redo a block, then resume — and only click **✓ Terminer et envoyer** (finish and send) once the session is genuinely done.
- Do I need a second screen?
- No, but it is more comfortable: the picture on one side, the band on the other. A television on HDMI does the job, and that is what most booths use.
- What if I fumble a line?
- During continuous recording, **↺ Refaire ce bloc** (redo this block) rewinds exactly to the start of the line under the playhead and stops there: you click to resume, and the new take replaces the old one in the edit, with no need to start from a blank band. To rehearse a scene before recording, the loop replays the same passage without hunting for it each time.
Density and reading
- Tracks are too far apart, I see few lines.
- The Density control in the band toolbar offers tight, normal or wide. Your choice is remembered.
Detection and adaptation
Adapters, dialogue writers · the service page
Offline detection
- Does lip detection need a connection?
- No, never. It is the feature that sets Axeho apart most: it consumes no external service and therefore costs nothing per use.
- What is a viseme, concretely?
- The shape the mouth takes for a sound. Two different sounds can produce the same image: that is what lets you change a word without the eye noticing.
Reading-rate control
- Is the rate flagged when a line is too short?
- Yes. A line that is too loose leaves a gap in the character’s mouth, and that fault costs as much as an overcrowded one.
Find and replace
- Can I replace a word for one character only?
- Yes. Pick the role in the band toolbar: search and replace are scoped to it, including “replace all”.
Transcription
Studios receiving an original version with no script · the service page
Twenty-six source languages
- Is Wolof supported?
- Yes, as are Yoruba, Lingala, Igbo, Hausa, Swahili, Xhosa, Zulu, Amharic, Shona, Somali, Northern Sotho and Ganda.
- What if the audio mixes two languages?
- Leave automatic detection on: it beats a forced choice on dialogue that switches language mid-scene.
Progress tracking
- How long does a transcription take?
- It depends on the media length and on your service. The application measures your previous runs and offers an estimate from the second one onwards, never before, so it promises nothing it has not observed.
- Can I switch tabs while it runs?
- Yes, the work continues. A sound signals completion, and can be turned off.
Transcribe with no network
- When is it better than the online service?
- When the content must not leave: embargo, confidentiality clause, a client who forbids it by contract. Or simply when there is no connection. These were jobs you had to turn down on principle.
- Is the result equivalent?
- No, and it is stated on screen before starting: the offline engine **does not separate speakers**. Every line arrives on a single character, to be split afterwards. Quality also depends on the model installed, and your own processor does the work.
Conforming
Studios facing a change of cut · the service page
Timecode-based carry-over
- What happens if a scene was cut?
- The lines involved are flagged as gone rather than silently deleted: it is your call whether the text moves elsewhere.
Series
Producers, project managers, dubbing directors · the service page
Series bible
- Can the bible be shared between adapters?
- Yes, by placing the projects folder on a shared space. Projects are small; only the audio stays on each machine.
Voices
Studios without floor time, mock-ups, previews · the service page
Your accounts, your costs
- Is voice generation included in the licence?
- No, and that is a choice. These services bill by usage: with your own accounts you see your volumes and choose your rates. A studio recording with human actors therefore pays for its licence alone.
- Does generation run on a server?
- No, on the workstation. In a team, adaptation is shared and generation spreads across machines.
Performance, measured on the original
- How is that different from analysing the text?
- “I’m fine” can be gasped, shouted or said flat. The text does not tell you; the signal does. The measurement compares each line to the average **of that same character**, a deep voice is not “dark”, it is normal for that actor.
- And when the measurement is not reliable?
- It stays silent. For a character with too few lines, no indication is issued rather than one drawn from two samples. Measured lines are shown next to the text: you see what the tool concluded, and you can overrule it.
Visual lip-sync: corporate and advertising
- Why rule out fiction?
- Because we sell an adaptation, not a fake. A retouched face passes once and always shows; written lip-sync holds up on a cinema screen. A corporate film or a spot has no such stake: the message comes first, and the sync gain is real.
- Where does the video go?
- To a third-party provider, the only Axeho function that does so. It is therefore absent until configured, and every upload asks for explicit consent, file by file. You can send a single shot rather than the whole episode.
Subtitles and delivery
Post-production, delivery managers · the service page
Seven import formats, five export formats
- My export is rejected by the client’s tool.
- Check the project frame rate: a DetX exported at 25 frames per second and read back at 24 shifts everything. It is set on the project, and timecodes are recomputed.
Broadcaster formats
- Why flag instead of correcting?
- Because a subtitle automatically shortened to fit a constraint is no longer your text, and nobody notices before broadcast. Axeho tells you which lines exceed the limit; the decision stays with the adapter.
- Does EBU STL support every language?
- No, and it is stated before export: its encoding carries neither Japanese, Chinese, Korean nor Thai. For those languages, delivery goes through TTML.
The file you can actually show
- What is it for, since mixing happens elsewhere?
- To get the work judged. A stem cannot be shown to a client, a production manager or a prospect: you need a file that plays. It is also the file you run through quality control before sending it.
Mastering to the broadcaster’s target
- How do I know the target was hit?
- Through quality control, which measures the final file and returns the deviation in figures. Mastering aims; control verifies. The two must not be the same action.
Quality control
Studios delivering to a broadcaster or a platform · the service page
Levels, on the final file
- Why show the deviation and not just “failed”?
- Because a verdict cannot be acted on. “−24.4 LUFS, target −23, deviation −1.4” tells you how much to make up; “failed” forces you to measure it again yourself.
- And if the file is silent?
- It is reported as silent, not as quiet. The difference matters: a loudness reading on a missing track returns a very low value, which reads as a level problem when in fact there is no sound at all.
Is every line actually there?
- How do you tell a hole from a simply quiet mix?
- The threshold is derived from the file itself, not fixed in advance. A deliberately quiet mix is therefore never declared empty, only a genuinely absent line is.
Offset and drift, told apart
- Why is drift so dangerous?
- Because it is zero at the start. You check the first minutes, everything lines up, and the end of the episode is unusable. It is the defect a play-through never finds.
- Does Axeho say where the drift comes from?
- When the slope matches a known ratio (24 taken for 23.976, a forgotten resample), the probable cause is named. It is a hypothesis, presented as such, but it is the hour you do not spend looking for it.
A sheet to attach
- Does Axeho fix what it finds?
- No, and that is deliberate. A check that fixes things by itself can no longer serve as evidence: you would not know what was right and what was patched. Mastering lives beside it, and is triggered by hand.
Licences and seats
- How many machines per licence?
- The seat count of your plan. Your desktop and your laptop count as two.
- I am changing computer : what do I do?
- Release the seat from the application, or remotely from your licence area with your key. A released seat becomes available again once its offline validation expires, thirty days at the latest.
- My computer died and I cannot release the seat.
- Write to us: we return the seat immediately.
- Do I need to be online to work?
- No. The application checks its licence whenever it has network, then runs for up to thirty days without a connection. Beyond that, it needs one check to resume.
- What happens when the year of updates ends?
- With a perpetual licence the application keeps working at the version reached. You lose neither the tool nor your projects; you simply stop receiving later versions.
- Can I install on a machine with no internet?
- The initial activation needs a connection, once. After that the machine works offline and revalidates whenever it finds network again.
Files and formats
- What can Axeho read?
- DetX (
.detx,.aae), SRT, VTT, TTAL (Netflix dialogue list), TTML, EBU STL,.xlsxand.xlsspreadsheets, and plain text. The detail is on the features page. - And write back?
- DetX, SRT, VTT, text and spreadsheet for the written side; EBU STL and TTML/IMSC for broadcasters and platforms. On the audio side: separate stems, timecode assembly, mastering to four targets. And the video render: the dubbed episode, picture and sound together, in a file you can actually show.
- Can I check what I deliver?
- Yes, and on the final file itself, video included: levels against the broadcaster’s target, the effective presence of every line in the mix, synchronisation offset and drift. The check produces a sheet to attach to the delivery and corrects nothing on its own. See Quality control.
- My videos are heavy and playback struggles.
- On import the application builds a light proxy and works on that, which is what lets it run on a modest machine. Timecodes still match the original video.
- Where are my projects on disk?
- In your user folder, separate from the application. An update never touches them. They are small, so they are easy to back up or to place on a shared space.
Adaptation and detection
- Does lip detection need a connection?
- No, never. The 77,288-entry lexicon is built in. That is what allows work in the booth, or on a programme that is not allowed to leave the studio.
- What is a “viseme”, concretely?
- The shape the mouth takes for a sound. Two different sounds can produce the same image, which is what lets you change a word without the eye noticing. Suggestions are filtered on that criterion, not on meaning alone.
- Does reading-rate control also flag lines that are too short?
- Yes, both ways. A line that is too loose leaves a gap in the character’s mouth, and that fault usually goes unnoticed until the recording floor.
- How does a series stay consistent?
- A bible keeps names, forms of address, recurring phrases, adaptation choices and voice casting from one episode to the next.
- The cut changed after the adaptation was written.
- The Conform module carries the existing text onto the new cut by timecode alignment, and flags what moved, disappeared or is new.
AI and voice services
- Why my own accounts?
- Because these services bill by usage, and the rights attached to a voice bind whoever uses it. With your own accounts you see your volumes, you choose your providers, and nothing goes through us.
- What happens with no accounts configured?
- Adaptation, lip detection, the rythmo band, alignment and exports all work normally. Only transcription and voice generation stay idle.
- Is my text used to train anything?
- Not by us: we never receive it. What goes to a provider leaves from your machine, on your account, under the terms you accepted with them.
- Which source languages for transcription?
- Twenty-six, including thirteen African languages: Wolof, Yoruba, Lingala, Igbo, Hausa, Swahili, Xhosa, Zulu, Amharic, Shona, Somali, Northern Sotho, Ganda. The full list is on the features page.
Teamwork
- How do we share a project?
- Put the projects folder on a shared space (Drive, OneDrive, Dropbox, NAS). Projects are small; the heavy audio stays on each machine.
- Two people on the same episode?
- The application warns without blocking, and refuses to overwrite someone else’s work without naming them.
- Can voice generation run on a server?
- No, it runs on the workstation. In a team, adaptation is shared and generation spreads across the machines.
When it does not work
- Windows refuses to install and mentions an “unknown publisher”.
- The application is not yet signed with a publisher certificate. Click “More info”, then “Run anyway”. The file is not at fault : the certificate is simply missing, and we say so before the purchase rather than after.
- The application asks for my key again although it was activated.
- Your machine has been offline for more than thirty days: it needs a connection, once, to revalidate. If the message mentions another computer, the hardware changed (motherboard, system reinstall): release the old seat from your licence area and activate again.
- “Licence already used on X seats”, although I released one.
- A released seat stays counted until its offline validation expires, thirty days at most. That is what prevents collecting activations by releasing them in a loop. If the machine is broken or stolen and you cannot wait, write to us: we return it straight away.
- Transcription or voices do not respond.
- First check your accounts under Settings → Keys and services. If the key is there, look at the balance or quota on your provider account: that is the most frequent cause, and it is not visible from the application.
- Adaptation suggestions never arrive, but detection works.
- That is the normal behaviour with no accounts configured: lip detection is local, suggestions need a connection and an account. Nothing is broken.
- The video will not play, or the waveform stays empty.
- Both light derivatives (audio and video proxy) are built on import, in the background. On a long episode, give it time to finish. If they are still missing, import the file again: some unusual containers do not expose their tracks.
- I cannot find a project.
- Projects live in your user folder, separate from the application. If you work as a team, check that the shared projects folder is the one you think it is: a project opened from a local folder does not appear in the shared folder, and the other way round.
- My export is rejected by the client’s tool.
- Check the project frame rate: a DetX exported at 25 frames per second and read back at 24 shifts everything. The frame rate is set on the project, and timecodes are recomputed.
Backups and data
- How do I back up my work?
- Copy the projects folder. Projects are small, a few hundred kilobytes, so a backup fits anywhere and is restored by copying it back. Audio and video stay wherever you keep them.
- Can an update erase my projects?
- No. The application and your data live in two separate places; an update only touches the application.
- What do you see of my work?
- Nothing. No video, no audio, no dubbing text is ever sent to us. We only know your licence key, a technical fingerprint of each activated machine, and the email address used at purchase.
Another question
Write to dubbing@djomionsolutions.com, mentioning your version (Help → About) and, if it is a fault, what you were doing when it happened.