Try the files. Check the result.
Camera footage, human speech and original test charts, with repeatable settings and measured outputs. Try a source before uploading your own media.
What is in the original test-chart pack?
Two four-second silent videos, a four-second reference tone, seven processor outputs, previews, a README and a manifest with file sizes, streams, settings and SHA-256 hashes.
You may use, copy, modify and redistribute the generated chart and tone fixtures and their outputs for any purpose. No attribution required. This permission does not apply to third-party software.
How these results were checked
Checked September 19, 2026. Generated four-second synthetic fixtures, executed Mediaivo commandFor()/runFfmpeg(), checked streams with ffprobe, decoded audio to mono 48 kHz float samples to measure sample peak and RMS. Not true-peak or LUFS measurements.
ffmpeg version 6.1.1-3ubuntu5 Copyright (c) 2000-2023 the FFmpeg developers
Synthetic chart/tone tests are not a benchmark for faces, speech, music, HDR, variable-frame-rate phone footage or every codec. No audio or video autoplays. Encoding versions may change byte counts.
The audio table describes decoded sample peak and RMS, not perceived loudness or true peak. Compression percentages describe only these sources. No overall quality score or ranking guarantee is implied.
Download the measurement manifest · Read the pack notes · FFmpeg rotation filter · FFmpeg volume filter
Camera footage and human speech
Download two prepared sources, four actual outputs, preview frames and a measurement manifest. Source credits, changes and reuse terms travel with the pack.
Executed the same commandFor/runFfmpeg functions as the Mediaivo tools. Checked file streams with ffprobe and decoded audio to 48 kHz mono float samples for sample peak and RMS. Not LUFS, true peak or a perceptual quality score. Encoder: ffmpeg version 6.1.1-3ubuntu5 Copyright (c) 2000-2023 the FFmpeg developers.
Compare compression on camera footage
This five-second camera clip shows a flower and a bee at 960 × 540. Compare moving detail as well as file size. It is one source, not a prediction for your video.
| Setting | Input bytes | Output bytes | Size reduction |
|---|---|---|---|
| Flower footage · Balanced | 964,911 | 243,705 | 74.7% |
| Flower footage · Smaller file | 964,911 | 122,658 | 87.3% |
Balanced uses CRF 25; Smaller file uses CRF 30. Both re-encode the picture. Smaller files can lose fine detail, and already compressed sources can grow. Preview the outputs before choosing.
Flower footage · Balanced — preview and repeat
Flower footage · Smaller file — preview and repeat
Source: Flower footage — MDN Web Docs contributors. CC BY-SA 2.5 under the MDN shared-assets content license. Audio and metadata removed; H.264 video stream retained. Outputs re-encoded using the indicated Mediaivo presets. Video derivatives and extracted poster frames are offered under CC BY-SA 2.5.
Processed with the same Mediaivo tool code. Measurements, settings and SHA-256 hashes · Credits and reuse terms. One short nature clip and one prepared speech excerpt. These results do not represent every phone, speaker, codec, recording or scene. No claim of lossless compression, noise removal, loudness normalization or endorsement by the source creators.
Hear gain changes on a human voice
A 12-second human narration excerpt, deliberately reduced to 35% amplitude to make a quiet input. Keep your device volume low and unchanged while comparing. Playback starts only when you press play.
| File | Peak (dBFS) | RMS (dBFS) | Full-scale samples |
|---|---|---|---|
| Quiet source | -15.44 | -31.42 | 0 |
| Speech · 125% | -13.57 | -29.48 | 0 |
| Speech · 200% | -9.39 | -25.4 | 0 |
125% adds about 1.94 dB; 200% adds about 6.02 dB before encoding. This example stays below full scale. Louder source recordings can clip: gain does not remove noise, repair distortion or normalize loudness.
Download input · Download output
Try Speech · 125% in the tool →
Download input · Download output
Try Speech · 200% in the tool →
Source: Alice’s Adventures in Wonderland, Chapter 1 — Lewis Carroll; read by Kristen McQuillin for LibriVox. Public-domain recording released by LibriVox. Excerpt 00:35–00:47, decoded to mono 48 kHz WAV, amplitude multiplied by 0.35 to prepare a quiet example. Outputs exported as MP3 at 125% and 200%.
Processed with the same Mediaivo tool code. Measurements, settings and SHA-256 hashes · Credits and reuse terms. One short nature clip and one prepared speech excerpt. These results do not represent every phone, speaker, codec, recording or scene. No claim of lossless compression, noise removal, loudness normalization or endorsement by the source creators.
See exactly what a 90° rotation changes
The green “TOP LEFT” and amber “BOTTOM RIGHT” markers make direction visible. This original, silent four-second chart is physically rotated clockwise, not just given a playback tag.
| Check | Input | Output |
|---|---|---|
| Dimensions | 640 × 360 | 360 × 640 |
| Duration | 4 seconds | 4 seconds |
| File size | 9,632 bytes | 7,483 bytes |
| Video codec | h264 | h264 |
Download input · Download 90° clockwise output
Try 90° clockwise in the tool →
Check direction, then check playback
The upper-left marker moves to the upper-right. A 180° rotation keeps width and height unchanged; 90° and 270° swap them. Mediaivo re-encodes the pixels to H.264 MP4, so this is not a lossless metadata-only edit. Phone videos with existing orientation tags can behave differently; preview your own result.
To change the frame shape instead, choose Crop Video to remove edges or Resize Video to fit the full picture with padding.
Measured with the actual Mediaivo processor. Settings, file hashes and encoder version · Test pack and method. Results can vary with the source and encoder version.
Compare 100%, 125% and 75% on the same audio
Start at a low device volume. This four-second synthetic reference tone is not speech or music. Keep your speaker or headphone setting unchanged while comparing the files; nothing plays automatically.
| File | Sample peak (dBFS) | RMS (dBFS) | Samples at or above full scale |
|---|---|---|---|
| Source WAV · 100% | -9.12 | -16.11 | 0 |
| Output MP3 · 125% | -7.18 | -14.17 | 0 |
| Output MP3 · 75% | -11.57 | -18.61 | 0 |
Download input · Download 125% amplitude output
Try 125% amplitude in the tool →
Download input · Download 75% amplitude output
Try 75% amplitude in the tool →
Gain is not loudness normalization
125% multiplies amplitude by 1.25 (about +1.94 dB); 75% applies 0.75 (about −2.50 dB). MP3 encoding can slightly change peaks. These references stay below full scale, but that does not prove every recording is distortion-free. The tool applies gain and exports MP3; it does not target a LUFS loudness, remove noise or repair clipping already in the source.
For an input peaking at −1 dBFS, adding +1.94 dB can push peaks above full scale. Listen to your loudest phrase before keeping an increase. Use Trim Audio for unwanted sections and Convert Audio when your main goal is a different format.
Measured with the actual Mediaivo processor. Settings, file hashes and encoder version · Test pack and method. Results can vary with the source and encoder version.
Balanced vs Smaller file: static and moving charts
Both inputs are original, silent four-second 640 × 360 H.264 clips. One has fixed markers; the other contains motion and detail. They are controlled test charts, not a benchmark for faces, sport or camera footage.
| Input / preset | Source bytes | Output bytes | Size change |
|---|---|---|---|
| Static chart · Balanced | 9,632 | 7,049 | 26.8% smaller |
| Static chart · Smaller file | 9,632 | 6,417 | 33.4% smaller |
| Moving chart · Balanced | 828,493 | 278,181 | 66.4% smaller |
| Moving chart · Smaller file | 828,493 | 165,005 | 80.1% smaller |
Both presets retain the canvas dimensions and nominal duration. Balanced uses CRF 25; Smaller file uses CRF 30. Neither promises an exact MB target or zero quality loss. Inspect edges and moving detail, not just the byte count. An already efficient input can even grow.
Static chart · Balanced — inspect and repeat
Static chart · Smaller file — inspect and repeat
Download input · Download Static chart · Smaller file output
Moving chart · Balanced — inspect and repeat
Moving chart · Smaller file — inspect and repeat
Download input · Download Moving chart · Smaller file output
Need a different canvas? Resize first. Need to remove footage? Trim it. For an estimated size based on duration and bitrate, use the Video File Size Calculator.
Measured with the actual Mediaivo processor. Settings, file hashes and encoder version · Test pack and method. Results can vary with the source and encoder version.
More examples already in Mediaivo
Try padded video resizing, a centered square crop and the SRT/VTT subtitle samples. Open a downloaded source in Metadata Viewer to compare its codec, dimensions and duration with the manifest. For a real recording, always inspect the result on the device and in the application that will receive it.
Writing about the examples? The resources and press page includes reusable links, source credits and a fact sheet.
October 10 processor checks
New original synthetic fixtures check one-interval trimming, white image padding, a JPEG crossing below 500 KB and fixed audio export settings. See the commands, measurements and hashes and CC0 reuse terms. These local processor checks do not establish subjective speech quality or physical-device compatibility.