Help

What is FFmpeg Video Studio?

FFmpeg Video Studio is a studio for cutting, grading and converting video, with ffmpeg as its engine. It has two views, chosen by the first two buttons of the header. The work view opens first: one video at a time, with a picture you can play, a cut, fades, colour grading, crop and rotate, texts and watermarks, and the sound with its own effects. The batch view converts many files at once: you add files or whole folders, choose how the video is to be encoded, and press Convert. You never type a command.

Work: one file at a time shows the work view.
Batch: convert many files with presets shows the batch view.

What sets it apart is that the choices are kept separate. The preset decides how the video is encoded; the size, the audio, the tags, the effects (a colour LUT is one of them) and the trim are each chosen on their own. One preset therefore serves every combination: you do not need a new preset for "the same, but smaller", "the same, but with a watermark" or "the same, but with the sound kept".

The rest of this page goes through what you can do: work on one video, convert a batch, let the app look after ffmpeg, choose a preset, limit the size, decide what happens to the sound, add effects, a colour LUT among them, choose which tags and chapters are kept and keep only part of each file.

The work view

The work view is for one video at a time: you see and hear what the conversion will make, set the cut and the fades on the timeline, and convert from there. Drop a video on the empty left side (Drop video files here to view or edit them) or add files and choose one in the list under the picture; a folder dropped here is ignored. When you switch to the batch view and back, nothing is lost: the list, the selection, the cut and the effects are kept, and a playing picture is paused on its frame.

Show the file list shows the file list again after a drawer has replaced it.
Hide the panel under the picture hides the panel under the picture, and Show the panel under the picture brings it back.

The picture, the sound and the timeline

The picture and its sound are played by ffmpeg with the effects you have switched on, so what you see and hear is what the conversion will make, apart from the final encoding. The picture is shown at most 960 pixels wide. The two are not locked to one clock, so in a long playback they can drift by a little. A file with no sound, or a computer with no sound output, plays the picture alone. Changing the file, the size or an effect while playing restarts from the same moment.

Under the picture the timeline is a film strip of 24 small pictures taken along the file, and under it a band that draws the sound at the same width, so a moment is at the same place on both. They show the file as it is, without the effects. A file with no sound has no band.

Play and Pause start and stop the playback; pausing shows the frame it reached.
Previous frame and Next frame move by one frame.
Volume is the loudness you listen at, and Mute and Unmute silence and restore it. It never changes the file.

Cutting

The buttons marked [ and ] set the start and the end of the part to keep at the position you are at. The part outside the cut is dimmed on the film strip. They write the same two boxes as Start and End (see Trim), so the marks and the boxes are one setting.

Start the cut here and End the cut here set the start and the end of the cut.
Go to the start of the cut and Go to the end of the cut move to the start and to the end of the cut.
Keep the whole file keeps the whole file.

Fades

A fade of the picture and a fade of the sound are drawn on the timeline, one over the film strip and one over the sound band. Each has a round handle at each end of the kept part: drag it inward to lengthen the fade, or type the seconds in the number box in the corner of the band. The picture fades to black; a length of zero is no fade. A fade stays inside the kept part and the two fades of a band never overlap.

The drawers and the effects of the work view

Under the picture a row of buttons chooses what the panel shows: the file list, or one drawer at a time. A drawer is the control of one effect, and using it switches that effect on. There are drawers for the colour grading, with bars or wheels, and the curves, and, for the sound, its channels, equaliser, compressor and level. Crop and rotate is one frame drawn over the picture, with handles to move, resize and turn it; texts and watermarks are boxes on the picture that you drag.

The work view has two effect lists of its own, Video Effects for the picture and Audio Effects for the sound. They are not the saved entries of the batch view. The standard effects are always there as rows with a switch, and the other effects are added from Add an effect, as in Adding an effect. Beside them the work view has its own choices for the file: preset, size, audio, tags, destination and what to do if the file exists; the panel writes to the selected files as you change it, so there is no Apply button here.

Effects listed in the work view on the ffmpeg tab of Settings chooses which effects the work view lists ready to switch on, for the picture and for the sound. The rest stay in the Add an effect menu, and an effect the app is built around is always listed. Back to the standard list goes back to the standard list.

The ffmpeg command

A panel under the file list shows the exact ffmpeg command for the selected file, with the panel's choices and effects, in a block you can select.

ffmpeg command is the panel, and Copy copies the command to the clipboard.

Convert a batch

This describes the batch view. Each row shows a small picture taken from the video itself, or an icon until it is ready.

Add files with Add files, or a whole folder with Add folder — every video file inside it is found, sub-folders included. You can also drag files and folders onto the window. Each file is a row in the list, with its name, its path and, once it is queued, a status. Clear the list empties the list except for what is running right now.

Press Convert and every file that is not already queued or running is converted, each with its own settings (see below). That includes files that were already converted, or failed, or were skipped: pressing it again converts them again, with the settings each row now holds. While a file is converting its row shows the percentage; hover over the status of a finished row to read where the result went, or, for a failure, the reason ffmpeg gave. The line at the bottom says how many files are ready, how far a run is, and how it ended — for example "3 converted · 1 skipped · 2 failed".

Every file carries its own settings — its preset, size, audio, tags, effects, cut, destination and what to do if the file exists. A file you add takes the choices the options panel shows at that moment, and only the newly added files are selected afterwards. The panel is a brush: it never shows the choices of a row, and selecting rows never changes it. To give other files different choices, select them, set the panel, and press Apply: it writes all of the panel's choices onto the selected rows that are not queued or running. It is off while nothing is selected, and while the panel's cut cannot be read. Select with a click, Ctrl (or Cmd) and click to add or remove a row, Shift and click for a range from the last click, or Ctrl+A for every row; a click on the empty part of the list clears the selection. The line at the bottom also says how many rows are selected.

Each row shows one small chip per choice — the preset, size, audio, tags, effects, destination and if-exists policy. A choice that leaves everything as it is (original size, original audio, all tags kept, no effects) is left out, so a plain row shows only its preset.

Apply — applies the panel's choices to the selected files.
Stop — appears while files are converting and cancels all of them. A cancelled file leaves no half-written output behind.
Cancel and Remove from the list — the cross at the end of a row cancels that file while it is running and removes it from the list once it is not.
Preset — how the video is encoded; see Presets. The button beside it opens the preset editor.
Files converted at the same time — how many files are converted at the same time, from one up to the number of physical cores in your computer, with a ceiling of four. The default is one, because a video encoder already uses every core by itself; more files at once only make them compete.
Open the output folder — opens the folder of the file that finished last, or the output folder you chose.

Where the converted files go

The destination drop-down of the options panel (Where the converted files are written) chooses where the converted files are written. Its first choice, Source Folder, writes each converted file beside its source and is built in; the other choices are destinations you made on the Destinations tab of the settings, which the button beside the drop-down (Edit the destinations) opens. There, as with presets, you add (New), copy (Duplicate) and delete (Delete) destinations and drag them into order; each has a name and a folder. The last choice of the drop-down, Pick Destination…, opens the system's folder picker: the folder you pick is saved as a destination named after the folder and chosen at once, and picking the same folder again chooses the same destination. Cancelling the picker changes nothing.

Folder — where the files are written. Leave it empty to write beside each file. The button beside it (Browse) picks a folder, and you can also type or paste a path.

A path that is complete (D:\Converted) is one place for everything. When you add a folder, its own name and its sub-folders are kept inside it, so two folders with files of the same name never land on top of each other.

A path that is relative — convert, .\convert or ..\convert — is counted from the folder of each file instead. .\convert is a convert folder beside the file; ..\convert is one beside the folder the file is in. This lets one setting serve files that sit in many different folders. The folder is created when it does not exist, and no structure is added to it: the folders the files already sit in are the structure.

The converted file is named after its source with the extension of the preset's container. It never replaces its source: if the name and the place would be the same, (converted) is added to the name.

What to do when the output file already exists — what to do when the output file already exists. Skip it (the default) leaves it alone and reports the file as skipped; Overwrite it writes over it; Keep both writes a second file, name (2), name (3) and so on.

ffmpeg, looked after

ffmpeg is the program that does the conversion. FFmpeg Video Studio uses the copy it downloads into its own folder, and never the ffmpeg that may already be on your computer, so it always runs a build it knows.

When none is installed, FFmpeg Video Studio installs it by itself the first time it starts. A small window shows how far it has got and closes when it is done. It fetches the ffmpeg yaiol has approved for its apps and checks it against the checksum yaiol published for it before it is unpacked into the folder all yaiol apps share. A download that does not match is thrown away. If it fails, the window says why and offers Try again.

The same copy serves every yaiol app, so the second app you open finds it already installed. Settings > General shows which version is installed, with a tick when it is the latest. When a newer version is approved, FFmpeg Video Studio installs it by itself at the next launch, in the same small window, which then says it is updating. If you ever need to do it by hand, Update is in Settings.

Why it is downloaded, not included

The ffmpeg builds that contain the best H.264 and H.265 encoders (x264 and x265) are licensed under the GPL. If FFmpeg Video Studio came with one inside it, the whole application would have to be GPL too. So the app does not carry ffmpeg: you fetch it, once, and it stays a separate program of its own.

The ffmpeg the app downloads has x264 and x265, so every preset works with it. The app never uses an ffmpeg you installed yourself, because it could not know what that one contains.

Presets: how the video is encoded

A preset says how the picture is written: the container, the video codec, how good it is to be, and how long the encoder may take over it. It says nothing about the picture's size, the sound, the colour or the effects — those are chosen next to it, see the sections that follow.

The preset editor is a page of the settings, opened by the button beside the preset drop-down. On the left is the list of presets: New, Duplicate and Delete work on it, and dragging a preset changes its place — the same order is used in the window's drop-down. On the right are the options of the selected preset; every change is saved at once, and there is no save button. A fresh installation comes with presets for the common cases: H.264, H.265, AV1, ProRes, uncompressed RGB and a straight copy.

Container and codec

The container is the file's outer form (.mkv, .mp4, .webm, .mov, .avi); the codec is how the picture inside is compressed. They are separate choices, and not every pair is possible: a container only offers the codecs it can hold. Choose the container first, and the list of codecs follows it. If you change the container to one that cannot hold the codec you had, the preset moves to one that it can.

As a rule: MKV holds almost everything and keeps subtitles; MP4 is the one every device plays; WebM is for the web and takes only VP9 or AV1; MOV is the editing and Apple world; AVI is old, and in practice here only for H.264 or uncompressed pictures.

Name — the name the preset has in the lists.
Container — MKV, MP4, WebM, MOV or AVI.
Video — the codec. H.264 and H.265 are the usual ones: widely played, small. AV1 and VP9 compress still further, but take longer. ProRes is an editing format: large files that are easy to edit. Uncompressed RGB writes the picture with no compression at all, which is lossless but enormous — about 9 GB a minute for full-HD video. Keep the original (no re-encoding) is described under copy and re-encode.

Quality or bitrate

An encoder is steered either by a quality number or by a target bitrate. With a quality number (CRF) you say how good the picture must be and the file gets the size it needs; lower is better and larger. The scale is 0 to 51 for H.264 and H.265, and 0 to 63 for AV1 and VP9. For H.264 around 18 to 23 is the usual range, and 20 is the default. With a target bitrate you say how large the file may be — so many kilobits for each second of video — and the quality is what that allows. It is done in one pass, so the result is less exact than a two-pass encode, which FFmpeg Video Studio does not have.

Use the quality number when the picture matters; use a bitrate when a file must fit a limit.

Encoding mode — Quality (CRF) or Target bitrate. Offered for H.264, H.265, AV1 and VP9.
Quality (CRF) — the quality number. On a graphics card (see Encoders) it is that card's own scale, which is not the CRF scale.
Bitrate — the target, in kilobits a second, when the mode is a bitrate.
Encoding speed — how long the encoder may take. Ultrafast (largest files) (only for H.264 and H.265), Fast, Balanced and Slow (smaller files): a slower setting makes a smaller file for the same quality. The AMD encoders have no speed setting.
Bit depth — 8 or 10 bits for each colour. Ten bits avoid visible bands in smooth gradients; it is offered for H.265 and AV1, and only plays on devices that support it.
Tune — tells H.264 or H.265 what the pictures are like: film, animation, grain, a still image, fast decoding or zero latency. Leave it at none unless one of these is exactly your material; H.265 has no film or still-image tune.
Extra x265 parameters — extra settings for the H.265 encoder, as key=value pairs joined by colons, passed to x265 as they are. For experts only.
ProRes profile — for ProRes only: 422 Proxy, 422 LT, 422, 422 HQ or 4444, from the lightest to the heaviest. ProRes has no quality number and no speed.

Copy, or encode again

Keep the original (no re-encoding) does not touch the picture: it is lifted out of the old container and put in the new one. That is very fast and loses nothing, so it is the right choice to change the container alone — an MKV into an MP4, say. Because the picture is not decoded, nothing that changes the picture can apply: the size and all the effects, the colour LUT among them, are ignored. The tags setting still applies, since it only concerns the file's tags and chapters. The sound, and the trim, work as usual.

Any other codec encodes the picture again, which takes time and, with a lossy codec, costs a little quality.

The encoder: software or a graphics card

Most graphics cards can encode video themselves, much faster than the processor. A preset names only the family of codec, H.264 or H.265; which engine encodes it is a separate choice, so one preset can be made on the processor or on a card.

Encode with on the ffmpeg tab of Settings is the default engine. Software gives the same file on every computer. Automatic takes the first graphics card that works, NVIDIA, then AMD, then Intel (Quick Sync), and the makers that work on your computer can be chosen one by one. Only the makers that passed a test on this computer are offered.
Encoder in a preset is Use default encoder, which follows the setting above, or Software, which always uses the processor.
Check the encoders again tests the cards again; a card or its driver can change without ffmpeg changing.

The app tests which cards work by encoding one small picture through each, when ffmpeg is found and again when you ask; the answer is kept until ffmpeg changes. A preset that a card cannot do stays on the processor rather than lose something silently: that is a 10-bit preset, a target bitrate, extra x265 parameters or a tune. A job started before the first test has finished is encoded on the processor too.

The price of the speed is the file size: at the same quality a card's encoder makes a larger file than x264 or x265, and its quality number is a different scale from CRF, so the same preset does not give the same file on another engine. Use a card when time matters more than size.

Size

The size limits how large the picture of the converted file may be. It is chosen in the window, apart from the preset, and edited under Sizes in the settings, with the same list tools as the presets. A fresh installation has the usual heights, from 4K down to 360p.

The picture is shrunk to fit inside the box you give, keeping its shape, and it is never enlarged: a 480p video converted with a 1080p size stays 480p. Leave a side empty — No limit — for no limit on that side. The remaining width or height is made even, which video encoders require.

Original size — the first choice in the window; leaves the size alone.
Maximum width and Maximum height — the box, in pixels.

A size does not apply to a copied picture.

Audio

What happens to the sound is chosen in the window, apart from the preset, from a list of audio entries that you edit under Audio in the settings. The list in the window offers only the entries that the chosen preset's container can hold: MKV takes AAC, Opus, AC-3, PCM or no sound; MP4 takes AAC and AC-3; MOV takes AAC, PCM and AC-3; WebM takes Opus; AVI takes AC-3, PCM and AAC. If you change to a preset that cannot hold the entry you had, the window shows Original audio.

Original audio

Original audio is not an entry of the settings but a rule, worked out for each file. It keeps the sound exactly as it is — not even decoded — when the container can hold that kind of sound. When it cannot, the sound is converted to what the container normally takes: AAC at 192 kbit/s, Opus at 128 for WebM, AC-3 for AVI. A ProRes picture always gets uncompressed PCM sound, which is what editing programs expect. A file with no sound is left without any.

This is the right choice most of the time. Choose a real entry when you want the sound changed.

An audio entry has these controls:

Audio — AAC, Opus, AC-3, PCM (uncompressed) or No audio.
Audio bitrate — for AAC, Opus and AC-3, in kilobits a second.

The channels, the level and the loudness are not settings of an audio entry: they are sound effects, kept in the Audio Effects list, and they work even when the picture is only copied. That list holds the same effects as the sound of FFmpeg Audio Studio, and is chosen in the options panel under What is done to the sound: an equaliser, a compressor, a noise gate, with Edit audio effects opening it. A preset or an audio entry from an earlier version that had a channel, level or loudness setting is converted to an Audio Effects entry automatically. A fresh installation has entries for AAC at three bitrates, AAC stereo, Opus, AC-3 stereo, PCM and no sound.

Effects

An effects entry is a named list of things done to the picture while it is converted. This list is called Video Effects; the sound has a second list of its own, Audio Effects, described under Audio. It is chosen in the options panel, apart from the preset, so one entry can go with any preset, and it is edited on the Video Effects tab of Settings.

Edit effects opens the effects tab.
New, Duplicate and Delete make, copy and remove an entry; dragging a row changes the order of the list.
Name is the name of an entry.
Add an effect adds an effect to the entry (see Adding an effect).
Close every effect closes every card of the entry, and the same button opens them again (Open every effect).
Drag to change when this effect runs is the grip on the left of a card; Remove this effect removes the effect from the entry.
Back to the default is the button beside a setting that puts it back to its default.

An effects entry

An effect is on while it is in the list, and off when you remove it; there is no separate switch. The same effect can be in the list more than once, each time with its own settings. On the page each effect is a card: a header that stays visible when the card is closed (the grip, the name of the effect, the button that removes it) and the settings below it. Under the settings the card shows the ffmpeg command it writes, so you can see exactly what will run.

Adding an effect

Add an effect is a category button joined to a search box. The category button offers All or one category and narrows the menu to it; typing in the search box narrows it by name. Choosing an effect adds it at the end of the list with its default values, and you drag it to the place you want. Besides the effects made for the app, the menu offers ffmpeg's own filters, each with the options ffmpeg gives it.

The order the effects run in

The effects run from the top of the list to the bottom, in the order you put them, so dragging a card can change the result. A few steps are placed by the app wherever they are in the list, because they must see the file at a certain point: for example the cut of a file, or the size limit. When an order cannot give the result you expect, because an effect counts on something that a later one changes, a line under the cards says so. The app never reorders your list.

All the effects

These are the effects the app offers, by category, with what each one does. The names and the sentences come from the effects catalogue that is installed together with ffmpeg, so they follow the version of ffmpeg you have; until ffmpeg is installed, the effects tab says that the effects are not installed yet.

Colour - Brightness, tint, white balance, curves

1D lookup table: Apply a 1D colour table from a .cube file.
Auto contrast: Stretch the picture so its darkest and lightest points use the full range.
Auto white balance: Remove an overall colour cast by assuming the scene averages to grey.
Black and white: Turn the picture grey, choosing which colours come out lighter.
Brightness, contrast, saturation, gamma: The four classic sliders in one filter.
Colour: Primary colour correction as in a grading panel: lift, gamma, gain and offset with a master and a red, green and blue bar each, plus temperature, tint, contrast around a pivot, saturation, hue and colour boost. Everything at its neutral value does nothing.
Colour balance: Push the shadows, mid-tones and highlights separately towards red, green or blue.
Colour contrast: Raise the contrast between the red, green and blue parts of the picture.
Colour curves: Reshape the tones with a curve for the whole picture and one for each of red, green and blue, as in a grading panel: drag a point of the line up or down to lighten or darken that part of the picture. A straight line from the bottom left to the top right does nothing.
Colour LUT: Apply a colour look stored in a .cube file.
Curves: Reshape the tones with a curve, as in a photo editor.
Equalise the histogram: Spread the tones evenly: a strong automatic contrast.
Exposure: Make the whole picture lighter or darker, like a camera's exposure.
Hue and saturation: Rotate the colours around the colour wheel, or make them stronger or weaker.
Hue, saturation and intensity by colour range: Like Hue and saturation, but for chosen colour ranges.
Levels: Set the black point and white point of each colour, as an image editor does.
Limit pixel values: Clamp every colour value into a range you choose.
Lookup table (RGB): Compute a colour table from a formula, on red, green and blue.
Lookup table (YUV): Compute a colour table from a formula, on brightness and colour planes.
Lookup table by expression: Compute a colour table from a formula.
Mix the colour channels: Mix red, green and blue into each other: sepia, swapped colours, tinted greys.
Selective colour: Change one range of colours (reds, blues...) without touching the others.
Shuffle colour planes: Swap the planes the picture is made of: strange, deliberate colours.
Swap U and V: Swap the two colour components: skin turns blue.
Table from two frames: Compute a colour table from two following frames.
Temporal equalisation: Even out the tones over several frames.
Tone mapping (HDR to normal): Convert high-dynamic-range video to standard range. Needs a correct HDR chain.
Vibrance: Boost the dull colours more than the strong ones: a gentler saturation.
Warmth: Make the picture warmer (orange) or cooler (blue).
White balance for shadows and highlights: Fix a colour cast in the darkest and lightest parts separately.

Sharpen and denoise - Crisper pictures, less noise and grain

Denoise (2D DCT): A frequency-based denoiser. Slow.
Denoise (3D FFT): A frequency-based denoiser working over time. Slow.
Denoise (non-local means): A careful denoiser. Slow.
Denoise (overcomplete wavelets): A wavelet denoiser. Very slow.
Denoise (wavelets): A wavelet denoiser with several settings.
Median filter: Remove isolated specks and keep the edges.
Median over time: Take the middle value of several frames to remove passing specks.
Reduce colour noise: Clean the coloured speckles and leave the detail.
Reduce noise: A fast, good denoiser for ordinary noisy video.
Reduce noise over time: Average several frames to calm the grain.
Remove grain: Take out film grain and small specks.
Sharpen: Make the picture crisper (or softer) with an unsharp mask.
Sharpen, adaptive: A sharpening that avoids making edges harsh.
Strong denoise (very slow): The most thorough noise remover of the list: block matching in three dimensions. Takes minutes per minute of video.

Blur and soften - Soft focus, smoothing, motion streaks

Average blur: A quick blur by averaging neighbouring pixels.
Blur, keep outlines: Soften the picture without blurring the edges.
Box blur: A simple, quick blur.
Directional blur: Blur in one direction: a motion streak.
Edge-preserving blur: Another blur that keeps the outlines sharp.
Gaussian blur: A smooth blur of a strength you choose.
Shape-adaptive blur: A blur that follows the shapes in the picture.
Smooth but keep the edges: Soften the picture while keeping the outlines sharp, as a skin-smoothing or cartoon look does.
Smooth, keep edges (bilateral): A smoothing that respects edges: skin softening, flat look.

Shape and position - Borders, rotation, perspective, zoom, upscaling

360 video projection: Convert 360-degree video between projections, or look into it.
Add borders: Surround the picture with a coloured border or bars.
Correct the perspective: Straighten a picture filmed at an angle by moving its four corners.
Crop: Cut a strip off any edge of the picture.
Crop and rotate: Turn the picture by any angle (a quarter or half turn is made exactly, with no empty corners) and cut a strip off each edge: one frame set over the picture. The turn comes first, so the corners it leaves empty are black unless the frame cuts them off.
Cut a picture into frames: Split one tiled picture into a sequence of pictures.
Fill the borders: Fill the edges of the picture, for example after stabilising.
Fit on a blurred background: Fit the picture inside a frame of another shape, and fill the rest with a blurred, enlarged copy of itself.
Flip: Mirror the picture.
Resize: Make the picture bigger or smaller.
Resize with colour-space control: A resizer that also converts the colour space and bit depth.
Rotate: Rotate the picture a quarter or a half turn, for a video shot with the phone on its side.
Rotate by any angle: Turn the picture by any angle, not only a quarter turn.
Scroll the picture: Move the picture past at a steady speed, like a ticker.
Slant: Shear the picture sideways or up and down.
Stereoscopic 3D: Convert between the ways a 3D picture is stored: side by side, over and under, one eye at a time.
Upscale pixel art (EPX): Enlarge pixel art by 2 with the EPX method.
Upscale pixel art (hqx): Enlarge pixel art by 2, 3 or 4 with the hqx method.
Upscale pixel art (Super2xSaI): Enlarge pixel art by 2.
Upscale pixel art (xBR): Enlarge pixel art by 2, 3 or 4 with the xBR method.
Zoom and pan: Slowly zoom into or across the picture. Made for stills: on video it needs care.

Time and frames - Frame tricks: time-lapse, smooth motion, padding

Change the frame rate, blended: Raise or lower the frame rate by blending neighbouring frames.
Drop duplicate frames: Remove frames that repeat the one before, for video that was stretched from a lower frame rate.
Drop near-duplicate frames: Remove frames that repeat the one before.
Fade: Fade the picture in from black at the start of the part that is kept, and out to black at its end. Each length is in seconds; 0 leaves that end as it is.
Frame rate: Force a constant number of pictures per second.
Keep one frame in N: Keep every Nth frame: a quick time-lapse look.
Pad the start or end: Hold the first or last frame, or add black, before or after. Changes the length.
Pick the best frame of each group: Keep one representative frame out of every group of frames.
Play backwards: Run the clip from its end to its start, sound included. The whole clip is held in memory.
Random frames: Show frames in a random order. A gimmick.
Repeat frames: Repeat a stretch of frames a number of times. Changes the length.
Shuffle the frames: Put the frames in an order you give.
Smooth motion: Invent the frames between frames, for smooth slow motion. Slow.
Speed (change timestamps): Play faster or slower by changing when each frame is shown. Changes the length.
Trim: Keep one section of the clip and drop the rest.

Fix problems - Banding, flicker, shake, interlacing, logos, lens

Apply telecine: Add a telecine pattern (film to video).
Correct lens distortion: Straighten the bulging of a wide-angle lens.
Deinterlace: Remove the comb pattern that interlaced video shows on moving things.
Deinterlace (bwdif): A good deinterlacer, an alternative to the built-in one.
Deinterlace (edge tracing): A deinterlacer that follows edges.
Deinterlace (kernel): A simple kernel-based deinterlacer.
Deinterlace (motion compensated): A deinterlacer that follows motion. Slow.
Deinterlace (neural network): A high-quality deinterlacer. Slow.
Deinterlace (Weston three-field): A deinterlacer using three fields.
Extract a field: Keep only the top or bottom field.
Field matching from hints: Match fields using a hints file.
Interlace over time: Combine frames into interlaced fields over time.
Interleave or deinterleave fields: Rearrange the lines of the fields.
Make interlaced: Turn progressive video into interlaced.
Pullup (undo telecine): Rebuild whole frames from a telecined field sequence.
Reduce dot crawl: Clean the dots and rainbows of old analogue video.
Remove a logo: Blend away a logo inside a rectangle you give.
Remove banding, quick: A fast way to smooth gradients.
Remove colour banding: Smooth the visible steps in skies and gradients.
Remove compression blocks: Soften the squares an over-compressed video shows.
Remove flicker: Even out brightness that pumps up and down between frames.
Remove judder: Remove the stutter that pulldown leaves.
Repeat fields: Apply the repeat-field flags of MPEG video.
Set the field order: Swap which field comes first.
Shift the field phase: Correct fields that are out of step.
Smooth compression artefacts (fast): A post-processing filter against block and ringing artefacts.
Smooth compression artefacts (pp7): Another post-processing filter against compression artefacts.
Smooth compression artefacts (simple): A simple post-processing filter against compression artefacts.
Smooth compression artefacts (slow): A slow, thorough post-processing filter.
Soften flashes: Reduce flashes that can trigger photosensitive epilepsy.
Split into fields: Split each frame into its two fields.
Steady shaky video: A quick, one-pass stabiliser.
Undo telecine (field matching): Rebuild the original film frames of a movie that was converted to 29.97 interlaced video.
Undo telecine (fixed pattern): Reverse a telecine pattern from film-to-video transfers.
Weave fields: Put pairs of fields back together into frames.
Weave fields, double: Weave fields into twice as many frames.

Looks and styles - Creative effects: negative, pixelate, glitch, trails

Amplify motion: Exaggerate what changes between frames.
Blend following frames: Combine each frame with the one before in a chosen way.
Colour channel offset: Shift red, green and blue apart: a glitch or 3D-glasses look.
Contact sheet: Tile several successive frames into one picture.
Custom matrix filter: Apply a matrix of numbers you give (blur, emboss, sharpen).
Dilate: Grow bright shapes with a structuring element.
Edges (Kirsch): An edge-detecting effect with a Kirsch operator.
Edges (Prewitt): An edge-detecting effect with a Prewitt operator.
Edges (Roberts): An edge-detecting effect with a Roberts cross operator.
Edges (Scharr): An edge-detecting effect with a Scharr operator.
Edges (Sobel): An edge-detecting effect with a Sobel operator.
Equation per pixel: Compute every pixel from a formula.
Erode: Shrink bright shapes with a structuring element.
False colour: Paint brightness levels in colours, like a thermal camera.
Film grain: Add noise or grain on top of the picture.
Flood fill: Fill a region of one colour with another.
Frequency-domain formula: Edit the picture's frequencies with expressions.
Grow bright areas: Make bright areas larger (a morphological effect).
Keep one colour, grey the rest: Everything turns grey except the colour you pick.
Keep one colour, grey the rest (HSV): Like the other colour-hold, picking the colour by hue.
Keep one colour, grey the rest (YUV): Like the other colour-hold, picking the colour differently.
Make a mask: Turn pixels into a black-and-white mask by thresholds.
Motion blur and ghosting: Mix several frames: smooth blur or echoes.
Motion trails: Leave ghost trails behind moving things.
Negative: Invert every colour, like a photographic negative.
Outline: Draw only the edges: a sketch look.
Pixelate: Turn the picture into big blocks.
Posterise: Reduce the picture to a few flat colours.
Scramble the pixels: Shuffle the pixels in blocks.
Shift the colour layer: Move the colour layer off the picture.
Shrink bright areas: Make bright areas smaller (a morphological effect).
Swap two rectangles: Exchange two areas of the picture.
Tilt-and-shift: A strange time-sliced look where time varies across the picture.
Tint with a colour: Give the whole picture a colour cast of your choosing.
Vignette: Darken the corners, as old lenses and films do.

Draw on the picture - Boxes, grids, vector drawings

Burn in subtitles (styled ASS): The same for styled ASS subtitles, with their own fonts and colours.
Burn in the subtitles: Draw the first subtitle track of the file into the picture. The output then has no subtitle track. Works with text subtitles only.
Draw a box: Draw a coloured rectangle, filled or only an outline.
Draw a grid: Draw a grid of lines over the picture.
Draw vector graphics: Draw shapes and paths from a vector description.
Text: Write text on the picture, as many as you like. Each one has a zone: the text is broken into lines to fit its width, and it can show the time played or left.
Text, every option: Write a line of text on the picture: a title, a name, a countdown.
Watermark: Put a picture, a logo for example, in a corner.

No effects

The first choice of the effects drop-down, No effects, is not an entry: it does nothing. It cannot be edited or deleted.

Tags: what the file says about itself

Besides the picture and the sound, a video file carries information about itself: who made it, when, where, with what. When a file is converted, all of it is copied across unless you say otherwise. A tags setting is a named choice of what is left out. It is chosen in the options panel, apart from the preset and the effects, and edited on the Tags tab of Settings. It works the same whether the video is re-encoded or only copied.

Tags is the drop-down of the options panel.
Edit tags opens the metadata tab.
New, Duplicate and Delete make, copy and remove a setting; dragging a row changes the order of the list.
Name is the name of a setting.
Tags chooses between keeping everything except what is listed, and removing everything except what is listed.
Exceptions: tags to delete (or Exceptions: tags to keep, in the other mode) is the list of tag names the mode is about.
Remove the chapters leaves the chapters out.

What a file carries

Information sits in three places.

On the file as a whole: a title, an artist, a comment, the date the file was made (creation_time) and, in a video from a phone, the place it was filmed (location, the GPS position) and often the device. Each is a tag, with a name and a value.
On each track, the picture, each sound and each subtitle: its language, sometimes a title of its own, a handler_name that the program that wrote the track left behind (it can name the device), and a creation date again.
In the chapters: named time ranges that a player shows as a skip bar. A chapter's name is its title.

Some fields are the container's own and not yours: ffmpeg's encoder tag, and technical fields such as the kind of MP4 it is. They are written again whatever you do, so the setting never names, removes or offers them.

Keep everything, or remove everything

Keep all tags copies every tag and removes only the names in the list. Use it to take out the GPS position and the date and leave the rest.

Delete all tags removes every tag and keeps only the names in the list. It is the safer choice when privacy is the point: a name nobody thought of is gone too, because it is not on your list.

With Keep all tags and location* and creation_time in the list, a phone video loses its GPS position and its dates, and keeps its title and the language of each track. With Delete all tags and title and language in the list, it keeps only those two.

Naming a tag

Write one name per line, or several on a line separated by ;. Case does not matter, and * stands for any run of characters: location* matches location and location-eng. While you type, chips under the box offer the names that cameras and phones usually write; a click puts the name in place of what you typed. Any name may be entered, including one only your camera uses.

A name applies everywhere it appears: on the file, on every track and on every chapter. So title in the list of names to keep keeps the title of the file, the titles of the tracks and the names of the chapters together.

Chapters

Remove the chapters takes the chapters out of the converted file. Without it the chapters stay, and what happens to their names follows the list like any other tag: with everything removed and nothing listed, the chapters stay but lose their names.

The starting setting

A fresh installation has one setting, Keep all tags, which changes nothing. It is an ordinary entry that you can edit, rename or delete, and there is always at least one: if the list is emptied, that one comes back.

Trim

The two fields Start and End under the options keep only part of each file. Type a time as seconds (90), minutes and seconds (1:30) or hours, minutes and seconds (1:02:03); a field left empty cuts nothing on that side. The same times apply to every file of the run.

The cut is not saved. It is forgotten when you close the app, so one left over from yesterday can never quietly shorten the files you convert tomorrow. If the end is not after the start, Convert is switched off and the line at the bottom tells you why.

When the picture is encoded again, the cut is exact. When it is only copied, a cut can only start at a keyframe — a picture the video can be started from — so it begins at the one just before your start time. The progress and the countdown of a text effect follow the part that is kept.

Other settings

The settings page, opened with Open settings, holds the lists described above and a few general settings. Help opens this help.

Language — the language of the app.
Theme — Light or Dark.
Show news from yaiol — whether FFmpeg Video Studio may show news from yaiol when it starts. It is one switch for every yaiol app on the computer; when it is off, nothing is requested at all.
Show notification when conversion is finished — on the Notifications tab. When a run ends, the system shows one notice, Conversion finished, with the same count the line at the bottom gives, for example "3 converted · 1 skipped · 2 failed". A run that was only cancelled shows nothing. It is on by default, and you can also mute FFmpeg Video Studio in your system's own notification settings.

When a newer version of FFmpeg Video Studio is out, a small news dialog opens on launch with a card that says New version available and the version, lists what is new since the one you run and offers Download. It only tells you: nothing is installed by itself, and a failed check is silent. Don't show again on a card keeps it from coming back; for the new-version card, until an even newer version ships. The new-version card does not depend on the news switch above.

Questions

A file fails with "Unknown encoder". Why?

The preset names an encoder that your ffmpeg does not have. The ffmpeg the app installs has the software encoders, and a graphics card is used only when it passed the test of the encoder setting, so this should not happen; if it does, choose another preset and check the encoders again in Settings.

Is there two-pass encoding?

Not yet. A target bitrate is done in a single pass, which is less exact than two. For a file of a given size, the best choice at present is a quality number, then checking the size.