Recording File Silent After a Driver Update, Picture Fine? Audio Source, Windows Audio Device and Driver Audio Component Diagnosis Guide

Key Summary
A recording with a perfect picture and no sound at all points at the audio capture chain. This guide uses three comparisons — playing another sound on the same machine, reading the file with MediaInfo, and watching the volume meter — to separate a source fault from a device fault, a track fault, and a driver audio component fault.
The Shape of the Problem: Picture Fine, Sound Missing
This guide covers one specific shape, and the shape itself is the whole diagnosis. A recording is made — with OBS, with NVIDIA ShadowPlay, with the Xbox Game Bar, or with AMD ReLive — and when the file is played back, the picture is exactly as expected: the game renders, the desktop is captured, the frames run smoothly. The sound is completely absent. Not quiet, not distorted, not dropping in and out — absent from the first second to the last. The file has a picture and no voice at all.
Three details define the shape, and all three have to be present. First, the picture is normal throughout. A recording whose picture is also broken belongs with a different guide. Second, the silence is total and constant: no game sound and no microphone sound, from the start of the file to the end. A file that starts with sound and loses it halfway through is a different shape. Third, the silence began at a identifiable point in time — very often just after a graphics driver update, a Windows update, or a change of headphones — and recordings made before that point, with the same settings, had sound. Without that before-and-after, there is no baseline, and without a baseline the diagnosis has no evidence.
That combination is what makes this symptom different from its neighbours. One neighbouring shape is a computer where no sound can be heard at all, live or recorded — no video plays with sound, no music plays, nothing comes out of the speakers. That shape is covered in the guide numbered 170, and it follows the live playback chain: the speakers at the end, the default device, the driver audio component, and the receiving device. This guide starts from the opposite side: the picture is fine and the question is whether live sound can be heard at all. Another neighbouring shape is a recording that stutters, where the encoder is overloaded and frames are dropped — that is covered in the guide numbered 171, and its shape is stuttering, not silence. A third neighbour is a recording that does have sound, but the sound crackles and breaks up when the machine is under load in a game — that is covered in the guide numbered 311, and its shape is degraded sound, not missing sound. A fourth neighbour is a finished edit whose sound is present but has changed pitch and speed — that is covered in the guide numbered 504, and the two guides are separated by one step in MediaInfo: no audio track at all follows this guide, an audio track with the wrong pitch follows that one. A fifth neighbour is the overlay that will not open at all — covered in the guide numbered 460, whose boundary section says in so many words that a recording that fails, or comes out silent, or produces no file, belongs to a separate topic further down the recording chain. This guide is the topic that section points at.
The most common first conclusion is that the driver update broke the recording. That conclusion jumps too far, and the honesty point has to be stated before anything else. "After a driver update" is a time correlation, not proof of causation. Outside sources say both things about driver updates, and both are true in different cases: an update can repair a damaged audio component, and an update can itself be the moment things went wrong — by moving the default output endpoint, or by overwriting an audio component. This guide never writes "the driver update broke the recording". It writes that the fault appeared at the same time as the update, which makes the update a variable to be compared, not a verdict. The comparison is simple: recordings made before the update, with the same settings, had sound; recordings made after the update, with the same settings, do not. Only then is the update worth suspecting — and even then, only one variable may be changed at a time, so that the result can be read.
One practical point has to be settled before anything else. A baseline must be written down before any test is run. The baseline is the recording tool and its version, the output settings, the file format, and the driver version, together with the answer to two questions: which sound is missing — all sound, only the game sound, or only the microphone — and when the silence started. Without that baseline, "silent after the update" is a feeling, and a feeling cannot separate a source fault from a track fault. The baseline should be written down, not remembered afterwards. Everything in this guide builds on it.
How Recording Splits Into Two Independent Paths
Recording is not one action. It is two parallel paths running at the same time: the picture path runs from the graphics card through rendering and the video encoder into the container file, and the sound path runs from the recording source through capture and the audio encoder into a separate track in the same container. The two paths are independent of each other. Either one can break while the other keeps going. A normal picture proves only that the picture path is intact. It proves nothing at all about the sound path — and that is the mirror image of the logic in the guide numbered 500, which says that a normal audio track proves only that the audio stream is intact, and says nothing about the picture.
The sound path itself divides into five segments, and any one of them can break while producing exactly the same surface symptom: a file with a picture and no sound. The segments are the recording source (which program's sound, and from which device), the Windows audio device (which endpoint the system sends sound to), the driver audio component (the part of the graphics driver that handles capture), the encoder track (how the sound is written into the file, and into which track), and the file container (how the file is finished and closed). "No sound" on its own cannot say which segment broke. The segments have to be compared one by one.
The most counter-intuitive segment is the driver audio component, because a graphics driver package contains more than one audio component, and the two that matter here do different jobs under similar names. On the NVIDIA side, besides the High Definition Audio component that carries sound out through HDMI or DisplayPort, there is a purely software virtual audio device — the NVIDIA Virtual Audio Device, known as NVVAD — which exists specifically for the capture and streaming audio path, for example the game audio in a ShadowPlay recording made through the in-game overlay. If that virtual device is disabled, or has disappeared from Device Manager, live listening can be completely normal while the recorded file has no game audio at all. Confusing the two components is, according to a guide on tutorialpedia.org, the number one reason a diagnosis of this kind stalls: the playback component and the capture component share a name and a driver package, but they are not the same thing, and testing one tells the reader nothing about the other.
Another segment that is easy to overlook is track mapping. OBS can record up to six independent audio tracks at once, and each track has to be assigned separately to decide which sources are written into it. An MP4 file, by default, stores only the first track, and many players only play the first track. If the recording output is set to record only the second track, while the desktop audio and the microphone sources are only ticked for the first track, the file will play back in complete silence — not because the sound was never recorded, but because it was recorded into a position the player never plays. Live streaming can be completely normal at the same time, because streaming uses a different track setting. That is why "the stream has sound but the recording does not" is not a contradiction. It is evidence, and it points at the track layer.
A third quiet failure belongs to the privacy and exclusive-mode segment. If the Windows microphone privacy switch is off, OBS will record silence without any error message, and the volume meter can still look normal. If another program has taken exclusive control of the microphone, the recording software may receive nothing at all. And a large Windows feature update can replace a working third-party audio driver with an incompatible version, after which the audio device has to be reinstalled and manually assigned again. None of these produce an error. All of them produce the same silent file.
The Deciding Forks: Three Comparisons That Narrow the Field
The whole guide turns on three forks. Each fork is a comparison, and each comparison removes a whole set of possibilities. They should be done in order, before any setting is changed, because a setting changed before a fork is tested destroys the evidence the fork would have given.
Fork One: Can the Same Machine Play Any Other Sound?
The first fork is the playback comparison, and it is the cheapest test in the guide. On the same computer, through the same speakers or headphones, play a different sound — an online video, a song, anything. If sound can be heard, the live playback chain is intact, and the fault is locked into the recording capture chain, which is what this guide follows. If no sound can be heard at all, the fault is in the playback chain, and the road is the one in the guide numbered 170 — the end speakers, the default device, the driver audio component, and the receiving device. This step costs almost nothing, and it directly decides which of the two roads the rest of the diagnosis follows, which is why it comes first.
Fork Two: Does the File Contain an Audio Track at All?
The second fork is the MediaInfo comparison. Open the recording file in MediaInfo and look at what tracks it contains. If there is no audio track at all, the sound was never written into the file, and the fault is in the capture, encoding, or container layer — it is not a player problem, and no player setting will bring the sound back. If there is an audio track but it plays back silent, note the number of tracks, the codec, and the sample rate, and move to the third fork. If MediaInfo cannot read the file at all, the file structure itself is damaged, which points at the corruption road described in the final section of this guide, not at a settings problem.
Fork Three: Does the Volume Meter Move?
The third fork is the volume meter comparison. Record a test clip of ten or fifteen seconds and watch the recording software's volume meter. If the meter moves but the file is silent, the sound reached the recording software and was lost at the step where it is written into the file — the track layer. If the meter does not move at all, the sound never reached the recording software — the source or device layer. This distinction is the second counter-intuitive core of this guide: a moving meter does not mean the file has sound. The meter only says the software received an input. The track mapping decides whether that input is ever written into the file. Outside sources agree on this point: a guide on screensnap.pro describes it as the most common shape of 2026 — "the recording runs smoothly, the meter shows green, but the saved file is silent" — and a guide on store.hollyland.com puts the same observation into a comparison table, where a meter with a reaction but a silent recording points at the track settings.
| Fork | Comparison | Result A | Result B |
|---|---|---|---|
| Playback comparison | Play another sound on the same machine, same speakers or headphones | Sound can be heard — recording capture chain, follow this guide | No sound at all — playback chain, follow the guide numbered 170 |
| MediaInfo comparison | Open the file in MediaInfo and list its tracks | No audio track — capture, encoding, or container layer | Audio track present but silent — move to the meter comparison |
| Volume meter comparison | Record a ten-second test clip and watch the meter | Meter moves but the file is silent — track layer | Meter does not move — source or device layer |
The Five Suspicion Layers
The five layers below are the whole map of this symptom. They are ordered by the cost of testing them, not by how likely they are, and the order itself is part of the diagnosis. The playback comparison is first because it costs almost nothing and it separates this guide from the guide numbered 170. The MediaInfo comparison is second because its result decides whether the problem is in the file itself. The meter comparison separates the source layer from the track layer, and the two sides lead to completely different next steps. The driver audio component layer is only worth suspecting after the first four layers have been excluded, or a driver will be reinstalled for nothing.
| Layer | What it is | Decisive test |
|---|---|---|
| First layer: recording source | The wrong source was selected, a source is muted in the mixer, the scene uses a different source, or the game is sending its sound to another device | Volume meter comparison: record a test clip and see whether the meter moves; if it does not, check each source in turn |
| Second layer: Windows audio device | The default playback device moved after a driver or Windows update, OBS is listening to "speakers" while the machine is actually playing through "headphones", or a Bluetooth headset switched | Device comparison: the device in the recording software's audio settings must be the same device the sound is actually heard through |
| Third layer: driver audio component | The NVIDIA Virtual Audio Device (NVVAD) is disabled or missing, the audio component was left unticked in a custom driver installation, or a Windows update overwrote the third-party audio driver | Device Manager comparison: look for NVIDIA, AMD, or Intel High Definition Audio and for NVVAD under Sound, video and game controllers, and check for a yellow warning |
| Fourth layer: encoder track | The track mapping is wrong, the MP4 stores only the first track, or the tracks ticked for a source do not match the tracks the output is recording | Track comparison: which tracks the output is set to record, against which tracks each source is ticked for in the advanced audio properties |
| Fifth layer: privacy, exclusive mode, and container | The Windows microphone privacy switch is off, another program has taken exclusive control of the microphone, or the MP4 container is at fault | Privacy comparison: whether microphone permission is switched on in Settings; exclusive-mode comparison: close other programs using the microphone and try again |
First Layer: Recording Source
This is the layer the meter comparison tests. When the meter does not move, the sound never reached the recording software, and the first place to look is the source itself: which device the desktop audio is set to capture, which device the microphone is set to capture, whether either source is muted in the mixer, and whether the scene in use actually contains the right sources. A game can also send its sound to a different device from the one the recording software is listening to — a USB headset, an HDMI monitor, a wireless headset, or a virtual reality headset are the cases a guide on geekchamp.com lists as the most common for this shape. The follow-up test is to check each source in turn, changing only one thing at a time, and recording a short test clip after each change.
Second Layer: Windows Audio Device
This is the layer the device comparison tests. After a driver update or a Windows update, the default playback device is the thing most likely to have moved. The recording software may be set to capture "speakers" while the machine is actually playing through "headphones", and then it will record silence — the sound exists, but it is travelling down a different path from the one being captured. The test is a direct comparison: the device named in the recording software's audio settings, against the device the sound is actually heard through. If they are not the same device, that is the layer. Bluetooth headsets deserve a special mention here, because switching to a Bluetooth headset can change the default device without any visible warning.
Third Layer: Driver Audio Component
This is the layer the Device Manager comparison tests. Open Device Manager, look under Sound, video and game controllers, and check for the NVIDIA High Definition Audio component, the AMD High Definition Audio Device, or Intel Display Audio, together with the NVIDIA Virtual Audio Device (NVVAD). A yellow warning, or a component that has disappeared entirely, is the signal for this layer. The component can go missing in three ways: it was disabled, it was left unticked during a custom installation of the driver package, or a Windows update overwrote it. Only when this layer has been reached — after the source, device, and track layers have been excluded — is it worth considering reinstalling the official graphics driver package, making sure the audio component is ticked in the custom installation options. Reinstalling earlier than that is guessing, and it cannot fix a source that was selected wrong or a track that was mapped wrong.
Fourth Layer: Encoder Track
This is the layer the track comparison tests, and it is the layer the outside sources argue about most. In OBS, the path is Settings, then Output, then Advanced, then Recording, to see which tracks are ticked for recording; and Edit, then Advanced Audio Properties, to see which tracks each source is ticked for. The tracks ticked for a source must line up with the tracks the output is recording. If they do not, the sound is recorded into a track the output never stores, or stored into a track the player never plays. For a test, use the first track first, because MP4 stores only the first track by default. ShadowPlay and Game Bar users do not have the same track settings, so their version of this step is different: check that the system sound and microphone switches are on, and check which device each one is capturing from. A guide on store.hollyland.com states the rule plainly: the recording output must tick the track number the source belongs to. This step is a comparison, not a recommended setting: find which track is missing first, and only then is the layer known.
Fifth Layer: Privacy, Exclusive Mode, and Container
This is the layer of silent switches. If the Windows microphone privacy switch is off, the recording software will quietly record silence, with no error message and a meter that can still look normal — a behaviour a guide on screensnap.pro calls out specifically. If another program has taken exclusive control of the microphone, closing the other programs and trying again is the comparison that tests it. And if the file itself is the problem rather than the sound — MediaInfo cannot read it, or the file was interrupted by a crash or a power cut while it was being written — then the fault is in the container, not in any setting. A guide on nasrtech.dev describes this road: a crash during recording can leave a file with a picture but no index, and a file that plays but has no sound can be the same kind of damage. That road is not this guide's settings road, and the boundary section below keeps the two apart.
The Seven Diagnostic Steps
Work through these steps in order. Each step is a comparison, and each comparison produces a written result. Change only one thing at a time. A step that changes two things at once produces a result that cannot be read, because no one can tell which change did the work.
| Step | What to do | What the result means |
|---|---|---|
| Step 1: Record the baseline | Write down the recording tool and version, which sound is missing, when the silence started, the output settings, the file format, and the driver version | Without a baseline, "silent after the update" is a guess, and the rest of the diagnosis has no evidence |
| Step 2: Playback comparison | On the same machine and speakers or headphones, play another sound | Sound can be heard means the recording capture chain, follow this guide; no sound at all means the playback chain, follow the guide numbered 170 |
| Step 3: MediaInfo comparison | Open the file in MediaInfo and list its tracks | No audio track means the capture, encoding, or container layer; an audio track that plays silent means move to step 6 |
| Step 4: Volume meter comparison | Record a ten-second test clip and watch the meter | Meter moves but the file is silent means the track layer, step 6; meter does not move means the source or device layer, step 5 |
| Step 5: Source and device comparison | Check each audio source, the mixer mute state, the privacy switch, exclusive mode, and the default playback device | The device in the recording software must match the device the sound is actually heard through |
| Step 6: Track comparison | Compare the tracks the output records against the tracks each source is ticked for | A source ticked for a track the output never records is the layer; test with the first track first |
| Step 7: Driver audio component comparison and stop | Check Device Manager for the audio components, then decide | A missing component or a yellow warning is the only signal for considering a driver reinstall; if the seven steps cannot separate the layers, stop and bring the records for inspection |
Step 1: Record the Baseline
The first step is to describe the fault precisely. Which recording tool is it — OBS, ShadowPlay, the Xbox Game Bar, or AMD ReLive? Is the whole file silent, or only the game sound, or only the microphone? Is the picture normal throughout? When did the silence start — after a driver update, after a Windows update, or after changing headphones? Write down the baseline: the recording tool and its version, the output settings, the file format, and the driver version. A baseline written down before testing is evidence. A baseline remembered afterwards is not. This step takes a few minutes and it protects every step that follows, because a diagnosis built on a guessed baseline is a guess all the way down.
Step 2: The Playback Comparison — The Decisive Test
On the same computer, through the same speakers or headphones, play another sound. If sound can be heard, the live playback chain is intact, the fault is locked into the recording capture chain, and the investigation moves to step three. If no sound can be heard at all, stop following this guide and follow the road in the guide numbered 170 instead. This is the decisive test of the whole guide, and it has to be run cleanly: the same machine, the same speakers or headphones, nothing else changed. It is also the step most often skipped, because it feels too simple to matter. Skipping it is how a recording fault gets treated as a playback fault, or the other way around, and every step after that is read against the wrong road.
Step 3: The MediaInfo Comparison
Open the recording file in MediaInfo and look at what tracks it contains. If there is no audio track at all, the sound was never written into the file, and the fault is in the capture, encoding, or container layer — move to step four. If there is an audio track but it plays back silent, note the number of tracks, the codec, and the sample rate, and move to step six, the track comparison. If MediaInfo cannot read the file at all, the file structure itself is damaged, and the fault belongs with the corruption road, not with a settings problem. The MediaInfo result has to be written down, because the number of tracks and the codec are the evidence that separates "never written" from "written but not played".
Step 4: The Volume Meter Comparison
Record a test clip of ten or fifteen seconds and watch the recording software's volume meter. If the desktop audio meter moves and the microphone meter moves, but the file is silent, the sound reached the software and was lost at the writing step — move to step six, the track layer. If the meter does not move at all, the sound never reached the software — move to step five, the source and device layer. A moving meter is not proof that the file has sound, and a still meter is not proof that the sound does not exist. Each result points at a different layer, and the two sides lead to completely different next steps, which is why this comparison has to be run before either side is investigated.
Step 5: The Source and Device Comparison
Check every audio source inside the recording software: which device the desktop audio is capturing, whether it is the same device the sound is actually heard through, which device the microphone is capturing, whether anything is muted in the mixer, and whether the scene in use contains the right sources. On the Windows side, check whether the microphone privacy switch is on, whether exclusive mode has let another program take the microphone, and whether the default playback device moved after the update. Change only one thing at a time, and record a short test clip after each change. A test clip that gains sound after one change proves that the change touched the layer the fault belongs to — and that is a comparison result, not a repair. The clip has sound; the question of why the setting moved is still open.
Step 6: The Track Comparison
In OBS, open Settings, then Output, then Advanced, then Recording, and look at which audio tracks are ticked for recording. Then open Edit, then Advanced Audio Properties, and look at which tracks each source is ticked for. The tracks ticked for a source must line up with the tracks the output is recording. For a test, use the first track first, because MP4 stores only the first track by default — a microphone placed on the second track will disappear in an ordinary player. ShadowPlay and Game Bar users should check instead that the system sound and microphone switches are on, and which device each one is capturing from. This step is a comparison, not a recommended setting: find which track is missing first, and only then is the layer known. Do not present a track setting as a fix. A recording that gains sound after the tracks are lined up proves the fault was in the track layer, and nothing more.
Step 7: Driver Audio Component Comparison and Stop
Open Device Manager and look under Sound, video and game controllers for the NVIDIA High Definition Audio component, the AMD High Definition Audio Device, or Intel Display Audio, together with the NVIDIA Virtual Audio Device (NVVAD). If there is a yellow warning, or a whole component is missing, only then is it worth considering reinstalling the official graphics driver package, making sure the audio component is ticked in the custom installation options. If all seven steps have been run and the layers still cannot be separated, or if the file comes with a crash, a power cut during recording, or a file MediaInfo cannot read at all, stop testing and bring the full record for inspection. And do not delete the original silent recording before the layers have been separated — it is the comparison material, and once it is gone, the before-and-after is gone with it.
Six Misconceptions That Mislead This Diagnosis
1. "I can hear the game, so the recording must have sound."
Wrong. Live playback and recording capture are two independent paths. Hearing the game only proves the playback chain is intact, and it proves nothing at all about the capture chain. The playback comparison and the MediaInfo comparison have to be run; a feeling that "heard means recorded" is not evidence.
2. "The volume meter moves, so the file must have sound."
Wrong. The meter only says the recording software received an input. When the track mapping is wrong, the sound can be lost at the step where it is written into the file. Outside sources agree on this: a moving meter with a silent file points at the track layer, not at the source layer. The meter and the file are two different measurements, and they have to be read separately.
3. "Updating the driver will definitely fix the recording."
Wrong. Updating a driver can be a repair, and it can also be the moment things went wrong — an update can move the default endpoint or overwrite an audio component. When a driver version is compared, the version number has to be written down, only one variable may be changed at a time, and updating the driver must not be presented as a universal fix. The sources disagree on this point precisely because the cause is different in each case, and the disagreement is written out in the section below.
4. "Reinstalling the driver fixes all recording problems."
Wrong. Reinstalling only fixes the driver audio component layer. A wrongly selected source, a wrongly mapped track, or a switched-off privacy switch will not be fixed by a reinstall. The guide numbered 170 states the same principle for the playback chain: reinstalling fixes only one link. Until the source, device, and track layers have been excluded, a reinstall is guessing.
5. "A silent recording means the file is broken, so buy a repair tool."
Wrong. If the sound was never written into the file, a repair tool cannot rebuild sound that was never there — the same limit the guide numbered 500 states for its own repair tools, and one the sources themselves state. The first question has to be whether the sound was "never written" or "written but not played". The first case needs a new recording. Only the second case is worth trying again. Buying a tool before that distinction is made is spending money on the wrong question.
6. "MP4 and MKV record sound the same way, so the format does not matter."
Wrong. MP4 stores only the first track by default, so a microphone placed on the second track will disappear in an ordinary player. For a test, use the first track first, or use MKV and split the tracks afterwards. Do not assume the format has no effect, and do not present changing the format as a universal fix. The format decides which tracks survive into the file, and that decision matters exactly when more than one track is being recorded.
Comparison Table: What Each Result Points At
| Observation | Most likely layer | Next test |
|---|---|---|
| Live sound can be heard, recording file is silent | Recording capture chain | MediaInfo comparison, step 3 |
| No sound can be heard at all, live or recorded | Playback chain | Follow the guide numbered 170 |
| MediaInfo shows no audio track | Capture, encoding, or container layer | Volume meter comparison, step 4 |
| MediaInfo shows an audio track, but it plays silent | Track layer or player layer | Track comparison, step 6 |
| Volume meter moves, file is silent | Encoder track layer | Track comparison, step 6 |
| Volume meter does not move | Source or device layer | Source and device comparison, step 5 |
| Recording software listens to "speakers", machine plays through "headphones" | Windows audio device layer | Device comparison, step 5 |
| NVVAD disabled or missing in Device Manager | Driver audio component layer | Driver audio component comparison, step 7 |
| Microphone privacy switch off, meter looks normal | Privacy layer | Privacy comparison, step 5 |
| File cannot be read by MediaInfo, or recording was interrupted by a crash | File corruption road | Stop and bring the record for inspection |
What the Sources Disagree About
The sources for this guide do not all say the same thing, and the disagreements have to be written out rather than settled silently.
Update the driver, or roll it back? A guide on geekchamp.com and a guide on store.hollyland.com say, on one side, that a damaged or outdated driver should be updated. A guide on geekchamp.com also says, on the other side, that when the problem appeared immediately after an update, it is worth checking whether other users have reported the same version, and in a small number of cases rolling the driver back or temporarily using an older version of the recording software. Both are right, because the cause is different in each case: updating helps when the driver is damaged or outdated, and rolling back helps when the update itself moved an endpoint or overwrote a component. A comparison has to record the version number and the time the fault appeared, and change only one variable at a time.
Which layer is "most common"? A guide on screensnap.pro says track mapping is the leading cause in 2026. A guide on store.hollyland.com puts a moving meter with a silent recording into the track layer in its comparison table. A guide on geekchamp.com puts selecting the wrong device first. The three are not in conflict: the symptom is different, so the leading cause is different — a moving meter points at the track layer, a still meter points at the device layer. This guide never says any one layer is "always the most common". The meter comparison decides which layer is followed, and the sources are quoted as the sources' own claims, not as measurements made for this guide.
MP4 or MKV? A guide on screensnap.pro says MP4 is not the right default for most Windows recording, because it stores only the first track, and a guide on store.hollyland.com also advises paying attention to the format. That is not a claim that MP4 itself is broken. It is a claim that, when more than one track is being recorded, the format decides which track survives. For a test, use the first track first. Do not present changing the format as a universal fix.
When to Stop Testing
Stop testing and bring the records for inspection if any of the following apply. All seven steps have been run and the layers still cannot be separated. The recording was interrupted by a crash, a driver crash, or a power cut, and MediaInfo cannot read the file at all — that is the corruption road, not a settings problem. The file has sound for the first half and loses it halfway through, which is a different shape from the total silence this guide covers. The sound is present but distorted, crackling, or breaking up under load, which belongs with the guide numbered 311. The overlay itself will not open, which belongs with the guide numbered 460. In every case, the written record is what makes the inspection useful: the recording tool and version, the output settings, the file format, the driver version, the MediaInfo track list, the meter result, and the result of every comparison, all with the date the silence started attached.
What to Bring for Inspection
Bring the computer, or at least bring the written record, and bring the numbers rather than the conclusions. The recording tool and its version, the output settings, and the file format. The driver version number, and whether the silence started after a driver update or a Windows update. The MediaInfo result — how many audio tracks the file has, which codec, and which sample rate. The result of the playback comparison, the volume meter comparison, the source and device comparison, and the track comparison, each with the exact step that was run. The published three-tier inspection and repair pricing for this service is HK$400–2,800, and the exact tier depends on what the inspection finds; no price is quoted before the inspection, because the fault has not been located yet. The shop is at Room 1209, Dali Plaza, Shek Mun. The telephone number is 2186 8010, and the WhatsApp number is 6558 6806. There is no shop in Mong Kok, and there is no on-site service.
Frequently Asked Questions
1. My recording has a perfect picture but no sound at all. Is the graphics card failing?
Not necessarily, and usually not. A normal picture with total silence points at the audio capture chain, not at the card's rendering. The picture path and the sound path are independent: the picture can be perfect while the sound path is broken at the source, the device, the driver audio component, or the track layer. Run the playback comparison first — play another sound on the same machine. If sound can be heard, the fault is in the recording capture chain, and the MediaInfo comparison is the next step. Do not condemn the hardware before both comparisons have been run.
2. I can hear the game while I play. Why is the recording silent?
Because hearing and recording are two different paths. Hearing the game only proves the live playback chain works. The recording capture chain — the source, the Windows device, the driver audio component, and the track mapping — is a separate path, and any one of its segments can break while playback stays normal. The NVIDIA Virtual Audio Device is a good example: it serves the capture path, and if it is disabled or missing, live listening is unaffected while the recorded file has no game audio. Run the MediaInfo comparison to see whether the file has an audio track at all, and then the meter comparison to see whether the sound ever reached the recording software.
3. The volume meter moves while I record, but the file is silent. What does that mean?
It means the sound reached the recording software and was lost at the step where it is written into the file — the track layer. The meter only says the software received an input. In OBS, the tracks ticked for a source have to line up with the tracks the output is recording, and MP4 stores only the first track by default, so a source ticked only for the second track will disappear in an ordinary player. Open Settings, then Output, then Advanced, then Recording, and compare the ticked tracks against the advanced audio properties for each source. A moving meter with a silent file is evidence, not a contradiction.
4. The silence started right after a driver update. Did the update break the recording?
It might have, and it might not have — the timing alone cannot prove it. An update can repair a damaged audio component, and it can also be the moment things went wrong, by moving the default output endpoint or overwriting an audio component. What matters is the comparison: recordings made before the update, with the same settings, had sound, and recordings made after the update, with the same settings, do not. Record the driver version number, change only one variable at a time, and do not present updating or rolling back the driver as a universal fix. Both directions are comparison tests, and the version number is the evidence.
5. How do I use MediaInfo to check whether a recording file has an audio track?
Open MediaInfo, drop the recording file into it, and look at what tracks are inside. No audio track at all points at the capture, encoding, or packaging layer — the sound was never written into the file. An audio track that is there but plays back silent means the sound was written in, so the next step is the track comparison. Write down the track count, the codec, and the sample rate. If MediaInfo cannot read the file at all, that points at file structure damage, not a settings problem.
6. What is the NVIDIA Virtual Audio Device, and why does it matter?
It is a purely software virtual audio device registered by the NVIDIA driver package, known as NVVAD, and it exists specifically for the capture and streaming audio path — for example, the game audio in a recording made through the in-game overlay. It is not the same as the High Definition Audio component that carries sound out through HDMI or DisplayPort. If NVVAD is disabled or has disappeared from Device Manager, live listening can be completely normal while the recorded file has no game audio. Check for it under Sound, video and game controllers in Device Manager, together with the High Definition Audio component, and look for a yellow warning.
7. Should I switch from MP4 to MKV?
Not as a first step, and not as a fix on its own. MP4 stores only the first track by default, so a microphone placed on the second track will disappear in an ordinary player — but that is a track mapping problem, and changing the format without fixing the mapping only moves the problem. For a test, use the first track first. If the tracks are lined up and the file still has no sound, the format is not the layer. The format decides which tracks survive into the file; it does not create sound that was never captured.
8. My microphone privacy switch was off. Could that be the whole problem?
It could be part of it, and it is worth checking early because it fails silently. If the Windows microphone privacy switch is off, the recording software will quietly record silence, with no error message, and the volume meter can still look normal. The same applies to exclusive mode: if another program has taken exclusive control of the microphone, the recording software may receive nothing. Check the microphone permission in Settings, close other programs using the microphone, and record a short test clip after each change, changing only one thing at a time.
9. Should I reinstall the driver?
Not first. Reinstalling only fixes the driver audio component layer. A wrongly selected source, a wrongly mapped track, a moved default device, or a switched-off privacy switch will not be fixed by a reinstall. Run the playback comparison, the MediaInfo comparison, the meter comparison, and the track comparison first. Only when Device Manager shows a yellow warning, or a whole audio component is missing, is a reinstall worth considering — and then it should be the official driver package, with the audio component ticked in the custom installation options. A reinstall is a comparison test, not a repair: a recording that gains sound afterwards proves the fault was in that layer, and nothing more.
10. When should I stop testing and bring it in for inspection?
When all seven steps have been run and the layer still cannot be told apart, or when the file comes with a crash, a power cut during recording, or MediaInfo cannot read the file at all. Bring the record: the baseline settings, the driver version number, the result of each comparison, and the MediaInfo record. Do not delete the original recording — it is the comparison material. The published three-tier inspection and repair pricing is HK$400–2,800, depending on what the inspection finds. The shop is at Room 1209, Dali Plaza, Shek Mun, telephone 2186 8010, WhatsApp 6558 6806.
Sources
- screensnap.pro, "OBS Not Recording Audio on Windows: 7 Fixes" — divides OBS audio into three layers, any one of which can break and leave silence; describes track mapping as the leading cause in 2026, with the shape "the recording runs smoothly, the meter shows green, but the saved file is silent"; states that MP4 stores only the first track by default, that a microphone on the second track disappears on playback, and that a switched-off microphone privacy switch makes OBS quietly record silence with no error message.
- geekchamp.com, "How to Fix Audio & Sound Not Working in OBS" — a moving meter with a silent recording points at the track settings; OBS set to capture "speakers" while the machine plays through "headphones" records silence, most commonly with USB headsets, HDMI monitors, virtual reality headsets, and wireless headsets; a mismatched sample rate can cause crackling and drift, and sometimes silence; when the problem appears immediately after an update, check whether other users have reported the same version, and in a small number of cases consider rolling the driver back or temporarily using an older version of OBS.
- store.hollyland.com, "OBS Audio Not Working: 8 Fixes" — a comparison table in which a meter with a reaction but a silent recording points at wrong track settings, and a stream with sound but a silent recording file points at wrong track assignment; a large Windows feature update can replace a working third-party audio driver with an incompatible version, after which the driver has to be reinstalled and the audio device manually assigned.
- tutorialpedia.org, "Fixing the NVIDIA Virtual Driver Problem on Windows 11" — besides the display driver, the NVIDIA driver package registers audio components, and confusing them is the leading reason a diagnosis stalls; the NVIDIA Virtual Audio Device (NVVAD) is a purely software audio endpoint for the capture and streaming audio path, and when it is disabled or missing, clips are recorded with no game audio; after a display driver update with no sound, the official guidance is essentially that the update may have moved the default endpoint, and selecting the correct device restores sound.
- avaide.com, "Why ShadowPlay Sometimes Doesn't Record Audio" — six causes for silent ShadowPlay recordings, including system sound and microphone settings, driver compatibility with old audio and display drivers, software conflicts where OBS, Discord, or a virtual audio splitter takes the audio channel, and Windows audio permissions; the listed steps include updating the audio and display drivers and closing other programs using audio.
- clippingdaily.com, "How to Clip on NVIDIA: ShadowPlay Setup" — the silent ShadowPlay column names selecting the wrong audio device, and notes that after an update the overlay settings can be reset or moved, and have to be switched back on.
- nasrtech.dev, "Game Recording Corrupted? NVIDIA, Game Bar and AMD" — a crash, a driver crash, or a restart during a game recording can leave a file with a picture but no index, so it cannot be played; on the question of whether a file that plays but has no sound is the same problem, the answer given is that it can be, because a crash can take the recording's sound settings with it. This is the file corruption road, not the settings road in this guide.
- displaycard-repair.com guide 170 — HDMI and DisplayPort picture with no sound on live playback, worked through eight steps along the playback chain.
- displaycard-repair.com guide 171 — live and recorded stutter and encoder overload, worked through three kinds of dropped frames.
- displaycard-repair.com guide 311 — audio crackling and breaking up under load in games, through interrupt latency and the audio buffer.
- displaycard-repair.com guide 460 — the overlay will not open, through the switch, the shortcut, the display mode, permissions, and the program; its boundary section states that a recording that fails, or comes out silent, or produces no file, belongs to a separate topic further down the recording chain.
- displaycard-repair.com guide 500 — an edited export with sound but no picture, and the MediaInfo method of reading tracks.
- displaycard-repair.com guide 504 — an edited export whose sound has changed pitch and speed, separated from this guide by the MediaInfo step: no audio track follows this guide, an audio track with the wrong pitch follows that one.