To compress a video for client review, export an H.264 MP4 from your master, keep the frame rate, and send a link instead of an attachment. Smol re-encodes a whole folder of review cuts on your Mac and, by default, writes copies and leaves the master alone. At its default (H.265, high quality) a 10-second 1080p clip went from 30,704,510 to 6,111,954 bytes.
Where the file is going decides the setting. Platforms publish recommended bitrates, transfer services publish caps, and a master should not be compressed at all. This table is the short version. The sections after it give the sources, the Smol settings, and the cases where another tool is the better choice.
| What you are sending | The limit that decides it | Start with |
|---|---|---|
| A review link (Frame.io, Vimeo, Dropbox) | Frame.io accepts files up to 5 TB and says compressing first is not needed. | The editor’s own export. To save upload time: MP4, H.264, “high”. |
| An email attachment | Personal Gmail and Outlook.com 25 MB. iCloud Mail and classic Outlook (IMAP or POP) 20 MB. Leave headroom. | MP4, H.264, “web”. Smol has no size target, so check the result against the limit. Details in the email guide. |
| A YouTube final | 256 GB or 12 hours. Recommended: 8 Mbps for 1080p at 24 to 30 fps. | Usually the editor’s export at 8 Mbps. Smol only for a slow uplink: MP4, H.264, “maximum”. |
| A Vimeo final | 10 to 20 Mbps for 1080p, and CRF 18 or below. | MP4, H.264, “maximum”. Check the bitrate on your own clip. |
| A download from WeTransfer or Dropbox | WeTransfer free: 3 GB in total (or 10 transfers) per rolling 30 days. Dropbox Basic: 2 GB per transfer. | Send the file as it is. If it is bigger: MP4, H.264, “high”, then check the size. |
| A master, ProRes or other intermediate | None. The master is the product. | Do not compress it. Keep it. |
What are the real upload and transfer limits for video?
Most limits are generous and a few are brutal. The generous ones belong to tools built for video. The brutal ones belong to anything that began as an email or a phone app. These were read on each service’s own page on 9 October 2026 unless the row says otherwise.
| Service | Limit | Source |
|---|---|---|
| YouTube | 256 GB or 12 hours, whichever is less. Uploads default to 15 minutes unless the account is verified. | YouTube Help, upload limits |
| Frame.io | Files up to 5 TB. Frame.io says there is no need to compress your media beforehand. | Frame.io Help |
| Dropbox Transfer | 2 GB on Basic, 50 GB on Plus and Family, 100 GB on Standard, Professional and Business, 250 GB on Business Plus and Enterprise. Links on Basic, Plus and Family expire after 7 days. | Dropbox Help |
| WeTransfer | Free: 10 transfers or 3 GB in total per rolling 30 days. Starter: 10 transfers or 300 GB in total. Ultimate: unlimited transfers up to 1 TB each. | WeTransfer plan limits, updated 30 May 2026 |
| 5 GB, 15 minutes, 256×144 to 4096×2304, 10 to 60 fps, 192 Kbps to 30 Mbps. | LinkedIn Help | |
| Instagram Reels (publishing API) | 300 MB, 3 seconds to 15 minutes, 1920 pixels wide at most, 25 Mbps VBR at most, 23 to 60 fps, HEVC or H.264. | Meta developer docs |
| TikTok (Content Posting API) | 4 GB, MP4 recommended (WebM and MOV accepted), H.264 recommended, 23 to 60 fps, 360 to 4096 pixels. | TikTok for Developers |
| Personal Gmail accounts and Outlook.com 25 MB (Workspace admins set their own limit). iCloud Mail 20 MB, or up to 5 GB with Mail Drop. Classic Outlook 20 MB by default with an IMAP or POP account, 10 MB with Exchange. | Gmail Help, Outlook.com, Apple Support and Microsoft Learn. More in our email guide. |
The Instagram and TikTok rows come from developer documentation for scheduling tools and APIs, and the in-app uploader can differ, so test one post first. Notice what is missing: a review platform with a cap that a short 1080p cut would hit. Instagram’s 300 MB is the closest, and a cut at YouTube’s 8 Mbps (about 60 MB a minute) passes it after about five minutes. The small numbers are in email and free transfer tiers.
Duration bites before size. At Vimeo’s 20 Mbps ceiling for 1080p, LinkedIn’s 15 minutes is about 2.25 GB, well under its 5 GB. For the audio side of a production, see compressing audio and video for podcasters.
What export settings do YouTube and Vimeo ask for?
| Setting | YouTube | Vimeo |
|---|---|---|
| Codec | MP4, H.264 High Profile, progressive, 4:2:0 | H.264 High Profile, Apple ProRes 422 (HQ) or H.265 |
| Bitrate, 720p | 5 Mbps (7.5 at 48 to 60 fps) | 5 to 10 Mbps |
| Bitrate, 1080p | 8 Mbps (12 at 48 to 60 fps) | 10 to 20 Mbps |
| Bitrate, 4K | 35 to 45 Mbps (53 to 68 at 48 to 60 fps) | 30 to 60 Mbps |
| Quality target | Bitrate only | CRF 18 or below |
| Frame rate | Match what you recorded | Native and constant; over 60 fps is reduced |
| Audio | AAC-LC, 48 kHz, 384 kbps for stereo | AAC-LC, 48 kHz, 320 kb/s CBR optional |
Sources: YouTube’s recommended upload encoding settings and Vimeo’s compression guidelines. Both are recommendations for what to upload, not targets for what to deliver to a client. Compress a final with Smol only when a slow uplink forces it; otherwise export from the editor at the published bitrate. Some upload endpoints transcode again, so a hard compression before upload can mean paying the quality cost twice. The only generation-loss measurement we have is an ffmpeg four-pass test, in the guide to reducing video file size.
The mismatch for Smol users is that these specs speak bitrate and Smol speaks CRF. CRF is a quality target, and the bitrate is whatever the footage needs. Here is what the quality labels produced on one clip, with the average bitrate worked out as bytes × 8 ÷ 10 seconds.
| Label | CRF | Bytes | Smaller by | Average bitrate |
|---|---|---|---|---|
| maximum | 18 | 15,432,056 | 49.7% | 12.3 Mbps |
| high (default) | 24 | 6,111,954 | 80.1% | 4.9 Mbps |
| balanced | 26 | 4,441,470 | 85.5% | 3.6 Mbps |
| web | 28 | 3,201,900 | 89.6% | 2.6 Mbps |
| low | 30 | 2,318,184 | 92.5% | 1.9 Mbps |
| tiny | 34 | 1,200,569 | 96.1% | 1.0 Mbps |
Measured with H.265 only, on Big Buck Bunny, an animated source with no audio track, so these bitrates are lower than camera footage will need. Handheld, noisy or fast footage lands higher at the same CRF. On this clip only “maximum” (VMAF 96.70) sits inside Vimeo’s 1080p range and above YouTube’s 8 Mbps. H.265 saved 18% to 30% against H.264 at matched quality on that clip, so expect H.264 files to land larger.
We have no Smol H.264 ladder, and CRF values do not carry across codecs. In the same post an H.264 encode at CRF 23 in ffmpeg scored VMAF 93.30 in 7,462,187 bytes, against 6,111,954 bytes at 93.83 for Smol’s H.265 default. For H.264, treat the labels as starting points and check your own result with size in megabytes × 8 ÷ seconds. We also have not documented what Smol does with video frame rate, audio codec or bitrate, or whether it moves the moov atom to the front of the file, so inspect one export in QuickTime Player’s Movie Inspector before relying on it for a platform spec.
For a hard bitrate or file-size target, Smol is the wrong tool: CRF cannot hit a byte count, and Smol’s video settings are codec, quality, speed preset, resolution and audio removal. Profile, GOP length and two-pass bitrate modes are not among them.
Which Smol settings fit each deliverable?
| Deliverable | Format and codec | Quality | Resolution | Notes |
|---|---|---|---|---|
| Review cut on a link | MP4, H.264 | high | Original | Smol’s labels were measured with H.265 only. Check an H.264 result on your own clip. |
| Review cut for a recent Mac or iPhone | MP4, H.265 (default) | high | Original | On the test clip, high scored VMAF 93.83 (above about 93 is hard to tell apart). All twelve H.265 outputs in our batch test carried the hvc1 tag, so QuickTime and Finder open them. |
| YouTube or Vimeo final | MP4, H.264 | maximum | Original | Vimeo asks for CRF 18 or below. We have not measured Smol’s H.264 at this label, so check the bitrate against the table above. |
| Email attachment | MP4, H.264 | web | 854 or 1280 | Smol cannot target a size, so the result can land over the limit: check it, and re-encode at a lower quality if it misses. Width choices are original, 3840, 1920, 1280 and 854. The email guide’s test favored 960×540 over 1280×720 on both clips, and Smol offers neither; 854 is the nearest lower choice and untested. Smol’s H.264 “web” label was not measured; the email guide’s one Smol H.264 check used CRF 23 on a 3-second cut. |
| 4K master to a 1080p copy | MP4, H.264 or H.265 | high | 1920 | Smol’s “high” was not tested on 4K. In the 4K guide’s ffmpeg test (x265, CRF 28, a 4-second synthetic 4K clip), the 1080p file was 13.4% of the 4K size and scored VMAF 99.94 on a 1080p display, but 90.04 on a 4K display. |
| Vertical social clip | MP4, H.264 | high | Original | Resolution values act as boxes and we have not tested a vertical clip, so test one before a batch. |
Two details deserve a second look. Smol’s resolution choices are bounding boxes: setting 1920 on a 3024×1964 recording produced 1662×1080. We have not tested a vertical clip, so check one before a batch. And there is no 960×540 option, which won a busy-footage test in the email guide. For screen recordings, see compressing a Mac screen recording: a 60-second, 56,384,195-byte recording became 3,691,160 bytes at Smol’s default and 1,399,221 at resolution 1920, in 16.3 seconds.
How do you compress a review cut, step by step?
Start from the master or the best export
Every compressed copy should come from the highest-quality file you have, never from a previous copy. In ffmpeg, four identical CRF 28 passes took a clip from VMAF 90.03 to 77.37 while the file shrank only 39%. The numbers are here.
Drop the files or the folder onto Smol
Video, images and PDFs can share one batch. By default each output is a copy next to the original with a
_compressedsuffix, so originals are not overwritten. Output location and naming can be changed in settings.Choose MP4 and the codec
Use the Compress operation, not Convert: quality is a Compress setting. H.264 for anything a client or a platform will open. H.265 is Smol’s default and is the right pick only for your own library or for recipients you know are on recent Apple devices.
Choose quality and resolution
“high” for review cuts, “maximum” for finals, an exact CRF if you have a number in mind. With H.264, check the result on your own clip. Leave resolution on original unless the destination is smaller. Switch audio removal on for silent b-roll and loops.
Run it and read the per-file results
Smol shows progress and savings for each file. On its direct compress path, if compressing would make a file larger, Smol keeps the original instead of writing a bigger copy.
Watch it before it leaves
Scrub to the busiest 30 seconds, read the smallest on-screen text, listen for audio sync, then divide the size by the duration. If it looks blocky, lower the CRF by 2 to 4 and re-encode from the source.
For a single file, the Finder Quick Action adds a “Compress with smol” item to Finder’s right-click menu. For a whole shoot, the batch guide compares the four routes on twelve clips.
What should a client review cut look like?
Small enough to open from a link on a laptop or a phone, sharp enough that the client can judge the edit. Resolution is your call, and Smol offers original, 3840, 1920, 1280 and 854 wide. Smol does not add a watermark or burned-in timecode, so do that in the editor before you export. Then open the compressed copy and the source in the same player and compare the darkest scene and the skin tones before anything leaves.
| Bitrate source | Average video bitrate | Megabytes per minute |
|---|---|---|
| Smol H.265 maximum, one 1080p clip | 12.3 Mbps | 92 |
| Smol H.265 high | 4.9 Mbps | 37 |
| Smol H.265 balanced | 3.6 Mbps | 27 |
| Smol H.265 web | 2.6 Mbps | 20 |
| Smol H.265 low | 1.9 Mbps | 14 |
| YouTube recommended, 1080p at 24 to 30 fps | 8 Mbps | 60 |
| Vimeo recommended, 1080p | 10 to 20 Mbps | 75 to 150 |
The Smol rows come from one animated clip, so camera footage will land higher. The point of the table is the order of magnitude: a ten-minute review cut at the high label is hundreds of megabytes on clean footage, not gigabytes. Check your own.
For where to put it, a review platform beats an attachment. YouTube says unlisted videos can be seen and shared by anyone with the link, so an unlisted link is not a private one. For unreleased work, that is a question for your client’s NDA, not for the encoder.
Should you deliver H.264 or H.265?
H.264 for anything a client or a platform will open. H.265 for your own library and for recipients you know are on recent Apple hardware. Smol defaults to H.265, so switching to H.264 is the first change to make for deliveries.
| Question | H.264 | H.265 |
|---|---|---|
| Full browser support (caniuse, read 9 Oct 2026: MPEG-4/H.264, HEVC) | 97.49% of usage | 19.99% (74.95% partial) |
| Size at CRF 28, calm pan | 1,050,171 bytes | 1,019,023 bytes (3.0% smaller) |
| Size at CRF 28, busy pan | 3,394,237 bytes | 2,380,623 bytes (29.9% smaller) |
| Platform stance | YouTube recommends it; Vimeo lists it first; TikTok recommends it | Accepted by Vimeo, Instagram’s API, TikTok’s API and Frame.io |
Our calm test clip sat near the 3% end and our busy one near the 30% end; your footage will land somewhere between or beyond. On calm footage like ours, codec does little for size, and the compatibility cost may not be worth paying. On a Mac, Safari plays HEVC everywhere, but our test found an MP4 tagged hev1 unplayable in AVFoundation while the same file tagged hvc1 played. The size and tag tests are in H.265 vs H.264: what actually changes.
How slow is software encoding, honestly?
Slower than your editor’s export by a wide margin, and that is the main reason not to use Smol on a deadline. Smol encodes in software and processes files one after another. Times below are from the sibling posts, on an Apple M2 Pro, with each post’s own conditions.
| Job | Software encode | Hardware or built-in encode |
|---|---|---|
| 10-second 1080p clip, 30,704,510 bytes | Smol default: 24.1 s, 6,111,954 bytes. Smol web: 13.8 s, 3,201,900 bytes. | QuickTime Export As 1080p: 2.2 s, 13,666,490 bytes, VMAF 81.96. |
| The same clip in HandBrake’s H.265 presets | Software preset: 32.0 s. | VideoToolbox preset: 3.8 s, 8,604,671 bytes, 33% larger than the software preset’s file. |
| Twelve 5-second 1080p clips, 135,034,271 bytes | A software loop in ffmpeg: 124.9 s. | ffmpeg with the hardware encoder: 15.9 s, and 3.08 times as much video as the software loop. |
| 4-second 4K ProRes clip, 218,034,032 bytes | An ffmpeg software encode: 45.7 s, 5,479,361 bytes. | An ffmpeg hardware encode: 4.9 s and 7.7 s, 7,412,303 bytes. |
| 60-second 3024×1964 screen recording | Smol default: 38.4 s. Smol at resolution 1920: 16.3 s. | Not measured. |
On our busy test Mac, Smol’s four-clip runs took anywhere from 24.06 to 85.99 seconds, so treat any single timing as a range. As arithmetic, not a measurement: 24.1 seconds per 10 seconds of 1080p, if it scales, is about 24 minutes for a ten-minute video.
What you get for the wait is quality per byte. Smol’s default scored VMAF 93.83 at under half QuickTime’s bytes. The 4K guide measured the same trade in ffmpeg: its hardware encoder needed 42.5 MB to match the quality its software encoder reached at 5.5 MB. Our view: for the 6 pm review cut, export with your editor’s hardware encoder and send it. Use Smol for the overnight folder, for finals going over a slow uplink, and for files other people sent you. The full trade is in the 4K guide.
Should you ever compress a master or an intermediate?
No. Masters and intermediates are large on purpose, because they have to survive grading, effects and a second export. Apple’s own ProRes figures put a 1080p, 29.97 fps ProRes 422 HQ stream at about 220 Mbps.
| ProRes flavor | Approximate rate | Ten minutes (derived) |
|---|---|---|
| 422 Proxy | 45 Mbps | 3.4 GB |
| 422 LT | 102 Mbps | 7.7 GB |
| 422 | 147 Mbps | 11.0 GB |
| 422 HQ | 220 Mbps | 16.5 GB |
| 4444 | 330 Mbps | 24.8 GB |
The other side, from the 4K guide: an ffmpeg software encode took a four-second ProRes 422 LT master of 218,034,032 bytes to a 5,479,361 byte delivery file. That file is right for watching and wrong for editing. Smol also writes only H.265 or H.264, so it cannot make a ProRes intermediate, and every pass through a delivery codec loses quality.
If only the container is wrong, do not re-encode at all. Rewrapping an H.264 MOV as MP4 with ffmpeg took 0.06 seconds and came out bit-identical in our MOV to MP4 test. Smol’s convert path always re-encodes, and on that clip the file grew by 37.3%.
How does this compare with HandBrake, Compressor and Media Encoder?
| Tool | Cost | Where it wins | Where it loses |
|---|---|---|---|
| Your editor’s export | Included with the editor | Straight from the timeline, so the file is encoded once. | Only for projects open in that editor, not for files other people send you. |
| QuickTime Player, Export As | Free | Two clicks and 2.2 seconds for a 10-second 1080p clip. | No bitrate or quality setting (Apple’s Export As guide). The 1080p export scored VMAF 81.96. |
| HandBrake | Free and open source | Filters, subtitles, chapters, a queue, bitrate modes and hardware encoders. Actively developed: 1.11.2 released 7 Jun 2026, commits through 9 Oct 2026. | A large settings surface to learn. Video only. |
| Apple Compressor | $49.99 on the Mac App Store | HEVC, H.264, ProRes and MXF output, HDR output, batch, droplets, folder watching. | Needs macOS 26.6 or later. Paid. |
| Adobe Media Encoder | A Creative Cloud app; the product page states no price | Ingest, transcode, proxies, presets and Watch Folders. | See Adobe’s plans page for what it costs. |
| ffmpeg | Free | Scriptable, and you choose every encoder setting. | A Terminal command per job. |
| Smol | $29 once | Drop a folder; video, images and PDFs in one batch; local; Finder Quick Action. | Software encode, no bitrate target, no filters, subtitles or chapters, macOS only. |
| Online compressors | Free or paid | No install. | The footage is uploaded to someone else’s servers. For unreleased client work, read the NDA first. |
HandBrake’s own documentation gives RF 20 to 24 as the range for 1080p, which is a useful starting point whatever tool you use. Adobe’s Media Encoder page lists ingest, proxies, presets and Watch Folders, and HandBrake’s repository shows commits through 9 October 2026. If you already own Media Encoder or Compressor, check them first for video work. See our HandBrake comparison for the long version. Smol earns its place when the folder is mixed, the editor is closed, and you would rather not learn an encoder’s settings.
Frame.io is a different case: it says you do not need to compress before upload and that it prepares your largest files for smooth playback. Take it at its word and send the export.
How does a studio adopt it, and what happens to unreleased footage?
Smol compresses on the Mac that holds the file. Nothing is uploaded, there is no account and no server, and it works offline. That describes what the software does. It is not a statement about any NDA or contract, which your client and your lawyer decide. Smol’s metadata stripping is documented for images and PDFs; we have not tested it on video files, so do not rely on it there. For stills from the same shoot, see removing EXIF data on a Mac. The wider question of handling client files is in Smol for confidential documents.
Pricing for a team is Solo $29 once for one Mac, Team $49 once for three Macs, and Enterprise $20 per seat, once, from 5 to 500 seats. A preset lives on one Mac, so write the settings down and recreate the same “review cut” preset on each Mac to get consistent output. The photos from the same shoot go through the same app; see Smol for photographers and Smol for marketing teams. Prices are on the pricing page.
A watched folder can compress whatever lands in a “Review exports” folder, with one caveat that matters here: watched folders can make files that are already optimized larger. Point one only at fresh, unoptimized exports, never at files you have already compressed. The behavior is described in the watched folder guide.
If an AI coding agent is part of your pipeline, Smol’s MCP server lets Claude Code, Codex or Google Antigravity run jobs locally. In the batch guide that meant telling the agent to compress a folder at a pinned CRF and reading the per-file byte counts back.
Where is Smol the wrong tool for video work?
Masters and intermediates. Smol writes H.265 or H.264 only. It cannot make ProRes, and it should never replace a master. If a colorist or another app needs an intermediate, use your editor, Compressor or Media Encoder.
A deadline. A hardware export finished a 10-second clip in 2.2 seconds where Smol took 24.1. If the review cut is due in an hour, export from the editor and send it. Larger files are the price, not a defect.
An exact bitrate or size. CRF targets quality, so Smol cannot land on 8 Mbps or 15.0 MB, and it does not expose profile, GOP or two-pass settings. Where a spec names those, use HandBrake’s average bitrate mode, ffmpeg or Compressor.
Subtitles, chapters, filters, deinterlacing or HDR work. HandBrake and Compressor do these; Smol does none.
A review platform that handles size for you. Frame.io takes files up to 5 TB without compression. Send the export.
A container swap. If an H.264 MOV only needs to be an MP4, ffmpeg -c copy is instant and lossless, and Smol would re-encode.
Not on a Mac. Smol is macOS only, and it needs macOS 13 or later on Apple Silicon, or macOS 14 or later on Intel.
Frequently asked questions
How do I compress a video for client review on a Mac?
Export an MP4 from your master, drop it into Smol, choose H.264 and original resolution, and send the _compressed copy as a link. Smol’s quality labels were measured with H.265 only: there, the default cut a 10-second 1080p clip from 30,704,510 to 6,111,954 bytes at VMAF 93.83. Check an H.264 result on your own clip, and keep the master untouched.
What are the best export settings for a YouTube upload?
YouTube recommends MP4 with H.264 High Profile and AAC-LC audio at 48 kHz, at 8 Mbps for 1080p at 24 to 30 fps (12 Mbps at 48 to 60 fps) and 35 to 45 Mbps for 4K. Keep the frame rate you recorded. Uploads can be 256 GB or 12 hours, whichever is less. Read on YouTube Help, 9 October 2026.
How do I send a large video file to a client?
Send a link, not an attachment. Frame.io accepts files up to 5 TB and says you need not compress first. Dropbox Transfer allows 2 GB on Basic and 100 GB on Standard. WeTransfer free allows 3 GB in total or 10 transfers per 30 days. Personal Gmail accounts and Outlook.com stop at 25 MB, so a video attachment must be small, and Smol has no size target to guarantee it.
Is H.265 or H.264 better for delivering video?
H.264 for clients and platforms. caniuse.com showed full browser support at 97.49% for H.264 and 19.99% for HEVC on 9 October 2026. H.265 saved 3.0% on a calm clip and 29.9% on a busy one at CRF 28. Use it for your own archive and Apple recipients.
Can I reduce video file size without losing quality?
Not pixel for pixel. A lossless re-encode of our test clip in ffmpeg was 236,974,886 bytes, 7.7 times larger. Smol’s maximum label (CRF 18, H.265) made it 49.7% smaller at VMAF 96.70, which is hard to tell apart from the source in normal viewing. Remuxing a container loses nothing but saves nothing.
Does Smol use the Mac’s hardware video encoder?
No. Smol encodes video in software, which is slower. A 10-second 1080p clip took 24.1 seconds at default and 13.8 seconds at web quality, against 2.2 seconds for QuickTime’s export, which produced larger files at a lower VMAF of 81.96. For deadline work, use your editor’s hardware export.
Should I compress my ProRes masters?
No. ProRes 422 HQ runs about 220 Mbps at 1080p, roughly 16.5 GB for ten minutes, because masters have to survive editing and grading. Compress copies for delivery, from the master, and archive the master first. Smol writes H.265 or H.264 only and cannot make a ProRes intermediate.
Keep reading
Guide
How to Compress a Video on Mac (4 Methods, Tested)
Guide
Reduce Video File Size on Mac Without Losing Quality
Video
How to Compress 4K Video on a Mac
Video
How to Batch Compress Videos on a Mac
Video
H.265 vs H.264: What Actually Changes
Comparison
HandBrake Alternatives for Mac (Without the Settings Maze)
Smol for other teams
