FFBlueprint — User Manual

FFBlueprint is a smart, zero-dependency command line builder for FFmpeg. It comes as a single, portable HTML file that works 100% offline in any browser. Please note: FFBlueprint only generates the text commands. You still need FFmpeg (and FFprobe) installed on your system to actually process your videos. It features deep hardware encoder support (NVIDIA NVENC, Intel QSV, AMD AMF, and Apple VideoToolbox) using their true native flags, smart container rules, and precise two-pass workflows. VideoToolbox requires macOS; its Apple Silicon quality mode requires FFmpeg 4.4 or newer, while Windows builds cannot provide it.

👉 Try FFBlueprint Live in your Browser

What sets this builder apart from other FFmpeg command generators:

• Four hardware families, with their real flags. AMD AMF, Intel QSV, NVIDIA NVENC (including their AV1 variants), and Apple VideoToolbox are covered. The PC GPU families use their native rate-control flags (-rc constqp -qp for NVENC, -qp_i/-qp_p/-qp_b for AMF, -global_quality for QSV); Apple uses VideoToolbox's native -q:v quality mode or -b:v bitrate mode instead of being forced into a PC-GPU control model. QSV's implicit-CBR quirk is handled for you.
• Quality boost checkbox. One click adds the flags that close most of the quality gap between PC GPU and CPU encoders: adaptive quantization (AQ), lookahead, multipass, B-frames-as-references — gated per encoder, since the AV1 variants don't take every flag. On NVIDIA it also switches CQP to quality-mode VBR (-cq), which behaves like CRF. Apple VideoToolbox uses its own native quality controls instead. See Hardware encoder options.
• Container selector with compatibility rules. MKV is the default; switching to MP4 renames the output extension, locks subtitles to mov_text, switches on Fast start, and warns about Opus/FLAC audio. Invalid container/codec combos are prevented instead of produced.
• Full subtitle handling. Drop, stream copy, codec conversion (mov_text for MP4, webvtt, srt, ass), or permanent burn-in with automatic path escaping — plus a stream selector (0:s:0). Most generators ignore subtitles entirely.
• Split by chapters. Generates the ffprobe chapter probe, parses the pasted start times, and builds the -f segment command with _001, _002… output names.
• Analyze Streams → stream pickers. Builds the ffprobe command for you; paste the output and every stream is listed in a table (codec, resolution, fps, pixel format, language, bitrate, duration). Parses JSON, multi-object JSON, and plain terminal text — and the results feed straight into the Audio Stream and Subtitle Stream dropdowns, so multi-track files (several languages, several subtitles) can be picked by their actual content instead of guessed indices.
• Correct two-pass workflow. Pass 1 drops the subtitle filter and audio, writes to the null device, and warns about the classic ffmpeg2pass-0.log pitfall. A "Copy Both Passes" button grabs both commands at once.
• Per-encoder everything. CRF ranges and defaults match the codec (x264 0–51/23, x265 default 28, VP9/AV1 0–63, SVT-AV1 default 35), and preset lists adapt: p1–p7 for NVENC, speed/balanced/quality for AMF, deadline + cpu-used for VP9, profiles for ProRes. No invalid preset/CRF combinations.
• Resolution done right. Fill only one side and the other becomes -2 (auto-calculated, kept even) — preventing the width not divisible by 2 failure.
• Single file, fully offline, zero dependencies. One FFBlueprint.html — no Node build, no internet connection, nothing phoning home. Windows-aware (NUL device, backslash paths, automatic quoting). On macOS, use a VideoToolbox-enabled FFmpeg build for the Apple encoders. Just open it, build, copy, run.

1. Open FFBlueprint.html in your browser (double-click it, or drag it into a browser window).
2. Fill in the sections top to bottom. The Generated Command box at the bottom updates live as you change anything.
3. Click Copy on the command box (or Copy Both Passes for two-pass encodes), paste into a terminal, and run.

Note about the input path: browsers don't expose full file paths, so dropping a file or using the file picker only fills in the filename. The field flashes orange when that happens — type or paste the full path in front of it (e.g. S:\Ai Videos\clip.mp4). Paths are quoted for you automatically.

Since the app can't run FFmpeg itself, this panel helps you inspect the file in two steps. You can skip it entirely — the command builds fine without it; the analysis just fills the Audio Stream and Subtitle Stream dropdowns so you don't have to guess track numbers:

1. Copy the shown ffprobe command and run it in a terminal against your file.
2. Paste the output into the text box. The app parses it (JSON or plain text) and lists every stream: index, type (video/audio/subtitle), codec, resolution, frame rate, pixel format, language, bitrate, and duration.

Use this to find e.g. which subtitle stream index to select, or whether your source is 8- or 10-bit.

• Software — runs on the CPU, best quality per bit, slowest:
  • libx264 (H.264, most compatible), libx265 (HEVC, smaller files)
  • libvpx-vp9, libaom-av1, libsvtav1 (AV1)
  • prores_ks (editing/intermediate codec)
  • copy — remux without re-encoding (instant, no quality loss, no size change)
• AMD AMF — h264_amf, hevc_amf, av1_amf
• Intel QSV — h264_qsv, hevc_qsv, av1_qsv
• NVIDIA NVENC — h264_nvenc, hevc_nvenc, av1_nvenc
• Apple VideoToolbox — h264_videotoolbox, hevc_videotoolbox (macOS only; requires an FFmpeg build compiled with VideoToolbox). On Apple Silicon, use its native Quality mode (-q:v, higher is better; FFmpeg 4.4+) or Bitrate mode (-b:v) as the compatibility fallback. VideoToolbox has no Quality Boost checkbox because its controls are different from NVENC/QSV/AMF.

Software encoders offer the following modes:

VideoToolbox shows its own two-mode selector: Quality (-q:v, Apple Silicon; higher values are better) or Bitrate (-b:v). VideoToolbox does not support the software two-pass workflow.

• CRF values — lower = better quality and bigger files. Ranges/defaults differ per codec (x264 0–51 default 23, x265 default 28, VP9/AV1 0–63, SVT-AV1 default 35). Roughly ±6 ≈ half/double the bitrate. For x265, start around 26–30.
• Preset — slower presets compress better at the same quality. medium is a good default; slow for archiving. The label changes per encoder (Deadline / CPU-used / Profile for VP9 / AV1 / ProRes).
• Tune — optional per-content tweak (film, animation, grain, …). Leave at none unless you know you need it.

AMF, QSV, and NVENC expose the PC-GPU controls below. Apple VideoToolbox is separate: choose Quality (-q:v) on Apple Silicon, where higher values mean better quality, or Bitrate (-b:v) as the compatibility fallback. It does not use the Quality Boost checkbox or the PC-GPU preset/rate-control flags.

• Rate Control — CQP (constant quality, like CRF), CBR (constant bitrate, for streaming/recording), VBR (average bitrate with optional peak cap).

• QP / Quality slider — used in CQP mode on AMF/QSV/NVENC. Lower QP is better; 20–26 is a sensible range for recording. For VideoToolbox Quality mode, the slider is -q:v and higher values are better; it requires Apple Silicon and FFmpeg 4.4 or newer.

• Quality / Preset — speed-vs-quality trade-off (speed…quality on AMF, p1…p7 on NVENC). VideoToolbox has no preset control.

• Bitrate / Max Rate / Buf Size — bitrate is used by all hardware families; Max Rate and Buf Size apply to AMF/QSV/NVENC CBR/VBR modes, not VideoToolbox.

• Quality boost — one checkbox that adds the flags which most improve quality per bit for the selected encoder family: adaptive quantization (-spatial-aq/-vbaq), lookahead, multipass, and B-frames-as-references. On NVIDIA, CQP additionally switches to quality-mode VBR (-rc vbr -cq … -b:v 0), which behaves like CRF (the bitrate floats to hit the chosen quality). Pair it with pixel format p010le (10-bit) for the full effect. Needs a recent GPU and driver — see Troubleshooting if a flag is rejected.

• Resolution — width × height. Blank = keep original. You can fill only one side: the other becomes -2, which auto-calculates and keeps the size even (odd dimensions break many players/codecs). Examples: 1920 × or × 1080.
• Frame Rate — common targets, or custom... for any value. Blank = keep original.
• Pixel Format — leave auto normally. Use yuv420p10le (software 10-bit) or p010le (hardware 10-bit) for HEVC/AV1 archival encodes; yuv420p fixes sources that some players can't decode (e.g. 4:4:4 or 4:2:2 x264).
• GOP Size (-g) — keyframe interval. Leave blank unless you have a specific requirement (e.g. 2×fps for scrubbing/editing).

• Codec — aac (default, universal), mp3, opus (best quality at low bitrates, WebM/own container), flac (lossless, no bitrate setting), ac3 (surround for home theater), copy (keep the original audio as-is), none (-an, strip audio).
• Stream — which audio track to keep, auto (first) by default (FFmpeg picks the track with the most channels). Run Analyze Streams and the dropdown lists every track with codec, channels, and language, e.g. 0:a:1 — aac · 6ch · lang: ger. Hidden when Map all streams is on (all tracks are kept anyway).
• Bitrate — 192k is a solid default for AAC; 128k is fine for speech.
• Channels / Sample Rate — usually keep. Set stereo/mono or 48 kHz when targeting a specific delivery spec.

Subtitle Stream — which subtitle track to copy/encode, default 0:s:0 (the first). Run Analyze Streams and the dropdown lists every subtitle track with its codec and language; custom... restores free-text entry (e.g. 0:s:1, or a full spec like 0:s:m:language:ger). Only used when Map all streams is off.

• Overwrite (-y) — don't ask before replacing the output file. On by default.
• Fast start (-movflags +faststart) — MP4 only: moves the index to the front so the file starts playing before it's fully downloaded. Managed by the Container selector: checked automatically for MP4, greyed out for MKV (where it does nothing).
• Map all streams (-map 0) — keep every stream (all audio tracks, all subtitles, chapters/data) instead of one of each. Hides the per-type Audio Stream and Stream Selector dropdowns, since individual picks don't apply.
• Split by chapters — see next section.
• Custom Arguments — appended verbatim at the end, e.g. -ss 00:01:00 -t 60 to trim, -metadata title="...", etc. Power-user escape hatch.

Cuts the output into one file per chapter:

1. Check Split by chapters. A third command box appears: the chapter probe.
2. Copy it, run it in a terminal — it prints one start time per line.
3. Paste all lines into the text box. The first line (0.000) is skipped automatically (the split points are the ends of each segment).
4. The main command gains -f segment -segment_times ... and the output name becomes name_001.mp4, name_002.mp4, …

Choosing 2-Pass shows two command boxes. They must be run in order, from the same folder:

1. Pass 1 analyzes the video and writes ffmpeg2pass-0.log into the current working directory. It produces no video file (-f null NUL) — that's normal, and it still takes a full encode's worth of time.
2. Pass 2 uses that log for precise rate control and writes the actual output.

If Pass 2 fails with unable to open file ffmpeg2pass-0.log: Pass 1 wasn't run, or it ran in a different folder. The log file is looked up relative to the directory you launch the command from, so cd into one folder and run both passes there. Use Copy Both Passes to grab them together. You can delete ffmpeg2pass-0.log (plus ffmpeg2pass-0.log.mbtree for x265) afterwards.

• Copy (top-right of each box) copies that single command.
• Copy Both Passes appears only in two-pass mode and copies pass 1 and pass 2 together, separated by a blank line.
• All paths and filter strings are quoted automatically; Windows-style backslash paths work as-is.

• Fast archive rip (GPU, NVIDIA): encoder hevc_nvenc, RC CQP, QP 24, Quality boost on, pixel format p010le, audio copy, Fast start on.
• Best-quality small file (CPU): encoder libx265, CRF 27, preset slow, pixel format yuv420p10le, audio opus (MKV is the default container, which Opus needs), audio bitrate 160k.
• Compatible web MP4: container MP4, encoder libx264, CRF 21, preset medium (Fast start switches on automatically), audio aac 192k, subtitles mov_text.
• Apple Silicon HEVC: container MKV or MP4, encoder hevc_videotoolbox, VideoToolbox Quality mode (-q:v, higher is better), audio aac or copy. Requires macOS, Apple Silicon, and FFmpeg 4.4+ with VideoToolbox support.
• Trim without re-encoding: Custom Arguments -ss 00:05:00 -to 00:15:00, encoder copy, audio copy.
• Fix a file that won't play on a TV: encoder copy (or x264 re-encode if the codec itself is unsupported), pixel format yuv420p.