VideoTech Video Converter

1. Quick start

1. Drag your videos onto the window — one file, several, or a whole folder.
2. Under Output options, pick what to convert to. MP4 (H.264) plays on everything.
3. Check the Destination folder.
4. Press Start.

That is the whole app. Everything below is detail.

2. Adding files

Drop files anywhere on the window, or use Add files… / Add folder…. Adding a folder picks up every media file inside it, including subfolders.

Each file is checked as it is added, so a corrupt or non-media file is refused straight away rather than failing halfway through a long batch. The list shows duration, size and resolution, and a thumbnail taken from 10% into the file.

Click a row to show that file in Explorer. Click ✕ to remove it.

3. Choosing a format

Video

If the video inside cannot legally go in the container you picked, the file is skipped with an explanation rather than being silently re-encoded.

Audio

MP3, AAC (M4A), Opus, FLAC, WAV, ALAC and Vorbis. Choosing an audio format for a video file extracts its sound.

4. Options

• Quality — Smaller file, Balanced, Best quality.
• Resolution — never upscales. Picking 1080p leaves a 720p file at 720p.
• Frame rate — never speeds anything up. Picking 60 leaves a 30 fps file at 30 fps.
• Audio — Convert re-encodes it, Keep original copies the existing track untouched where the container allows it, Remove drops sound entirely.

One set of options applies to every file in the list. If you need different settings for different files, run a second batch.

5. Automatic subtitles

VideoTech can listen to the speech in your video and write subtitles from it. This happens entirely on your computer — nothing is uploaded, and it works with no internet connection once the model is downloaded.

Under Output options → Subtitles:

Translate to English transcribes speech in another language straight into English subtitles.

Transcription happens before the conversion and can take longer than the conversion itself. The progress bar covers both stages, and the row shows Transcribing… while it works.

A few things worth knowing

• If there is no speech, the file is converted normally and the row says so.
• Accuracy depends on the model you downloaded and on how clear the audio is. Background music and several people talking at once are the hard cases — turning on Clean up audio helps considerably.
• Even when you burn subtitles in or embed them, the .srt is kept beside the output so you can correct it and run the file again.

6. Clean up audio

Removes the background noise from speech and leaves the voice: hiss, hum, traffic, air conditioning, a fan, room tone. It is a speech-enhancement model, not an equaliser or a noise gate, so it does not simply turn quiet parts down — it separates the voice from everything else.

Under Output options → Clean up audio:

Stereo stays stereo — each channel is processed separately.

It is quick — this is audio, not video, so it adds seconds to a job rather than minutes.

7. Remove silence

Cuts the parts where nobody is speaking. It finds speech with a model rather than by measuring volume, which is the difference between this and the “silence remover” in most tools: room tone, a fan, breathing and a distant door are all loud enough to fool a volume threshold, and a word trailing off at the end of a sentence often is not loud enough to survive one.

Under Output options → Remove silence:

A little padding is always left on both sides of each phrase, because cutting exactly on the boundary clips the start of the first word and the end of the last.

When the row finishes it tells you what it did — Removed 2m 14s of silence. If there was nothing worth cutting, the file is converted normally.

This works alongside subtitles: the captions are moved onto the shortened timeline for you, so they still line up after the cut.

8. Blur faces

Finds faces and blurs them, on every frame. Useful for anything filmed in public that you intend to publish — a street scene, a conference, a classroom.

Turn it on under Output options → Blur faces. There is nothing to position and nothing to keyframe; the detector runs on every frame, and a face that briefly turns away or is missed for a moment stays blurred for a short while afterwards rather than flickering back into view.

Blurring is per-frame work, so a conversion with it switched on takes noticeably longer than a plain one.

9. AI upscale

Rebuilds the picture at four times the size using Real-ESRGAN, instead of stretching it. On old phone footage, downloaded clips and screen recordings the difference is large — it reconstructs edges and texture rather than smearing the pixels it already had.

Turn it on under Output options → AI upscale. Two things to know before you do:

• It only accepts small sources — up to about 480p, which becomes 1080p. Four times a larger frame is past 4K, which nobody needs, so bigger files are skipped with a message saying so.
• It is slow, and how slow depends entirely on your graphics card. The model runs on a DirectX 12 GPU when there is one, and on the processor when there is not — and the difference is minutes per minute of video versus hours. The job row says Upscaling on the CPU — this is slow when it has fallen back. Try a short clip first and watch the estimate before committing to a long one.

To upscale and then cap the result — 480p up to 1080p rather than all the way to 1920p — set Resolution as well. The upscale runs first and the resolution setting trims the result.

10. AI models

The AI models page lists the optional downloads. They are separate downloads only because they are large files — everything is included with the app; there is nothing extra to buy.

Each feature names the model it needs and offers a button to fetch it:

Long clips are the thing to be careful with, not big ones: the cost is per frame, so a source that is small but long is just as slow as a large one. Start short and watch the estimate.

For subtitles, larger Whisper models are more accurate and slower; Tiny is faster and rougher. When more than one model of the same kind is installed, the most accurate one is used.

Each download is checked against a known size and fingerprint before it is accepted, so a half-finished or substituted file can never be used by mistake. If you already have a model file — from another PC, or because this machine has no internet — use Import a file….

Remove deletes a downloaded model. The app keeps working; only the features that needed it turn back off.

11. Where files are saved

Set a Destination folder, or tick Save next to each source file instead.

If a file with that name already exists, a number is added by default. You can change that in Settings to overwrite or skip instead.

12. While it runs

Each row shows a percentage and an estimated time. The window title bar and the taskbar button show the overall progress, so you can leave it running behind other windows.

Cancel stops the current file and deletes the half-written output. Closing the window cancels everything.

When done can open the destination folder, close the app, sleep, or shut down. Sleep and shutdown wait 60 seconds first — if you are back at the keyboard, open a Command Prompt and run shutdown /a to stop it.

The bottom-right corner shows how hard the converter is working and which encoder it is using — your graphics card, or the one built into Windows.

13. Settings

• Video encoder — leave on Auto. It picks the fastest encoder your machine can actually run, testing each one rather than trusting what it claims to support. Override it only when chasing a specific quality or compatibility problem.
• Convert several files at once — leave at one. Running several at once makes them compete for the same graphics hardware and usually finishes later, not sooner. Raise it only for AV1 or VP9, which encode on the processor.
• When a file with that name already exists — add a number, overwrite, or skip.
• Show thumbnails — turn off if you work with very large batches.

14. Command line

Everything the window does is scriptable:

VideoTech.exe --headless --to mp4_h264 --out D:\Converted D:\Footage

Options:

Format names are the ones in the dropdown: mp4_h264, mp4_hevc, mp4_av1, webm_vp9, mov_prores, gif, mp3, m4a, flac, wav, remux_mkv and the rest. The exit code is 0 when every file converted.

15. Where your files live

16. Troubleshooting

A file was skipped

The row says why, on both the Convert page and the Queue page. The usual causes are asking for audio from a silent file, remuxing into a container that cannot hold that video, or asking for an AI option the file cannot take — upscaling something already larger than 480p, or cutting silence while copying the streams untouched.

Conversion is slow

Check the encoder shown in the bottom-right. If it says GPU not used, your machine has no hardware encoder available and the work is being done by the processor. AV1, VP9 and ProRes are slow by nature.

Subtitles are wrong or missing

Quiet, noisy or overlapping speech is the hard case. A larger model helps, and so does turning on Clean up audio. If there was no speech at all, the row says so.

The subtitle option is greyed out

The speech model is not downloaded — the AI models page has it. The same is true of audio cleanup, remove-silence, face blur and upscaling: each one shows the model it needs and a button to get it.

Upscaling is taking hours

It has almost certainly fallen back to the processor — the job row says so while it runs. The model needs a DirectX 12 graphics card to be quick.

The cleaned audio sounds hollow or clipped

Use a lower strength. Remove as much as possible is aimed at bad recordings; on something already clean it can take the room out with the noise.

A face was missed

No detector is perfect, and the accurate detector model catches more than the fast one. Check the output before publishing.

Output is bigger than the original

You re-encoded something that was already compressed. Either lower the quality, or use Change container only if you only needed a different wrapper.

17. Known limitations

• No trimming, joining or editing — this is a converter. (Use BackendStudio for editing.)
• One set of options per batch, by design.
• Subtitles are generated from speech only; existing bitmap subtitles (Blu-ray/DVD) cannot yet be turned into text.
• No video stabilisation, and no automatic detection of black bars.
• Face blur has no preview: you see the result by opening the finished file.
• AI upscaling is fixed at 4× and limited to sources up to about 480p.
• Subtitle burn-in uses the default styling; fonts and colours are not yet adjustable.