When remuxing an AI short with FFmpeg, treat container metadata, chapters, and stream tags as three separate delivery decisions. Choose which values to copy from the input, which to set explicitly for the output, and which to leave out. Then inspect the output at each level: a successful remux does not prove that every requested field survived or was written as intended.
Separate the three layers
Container-level metadata describes the file as a whole, for example a title or author field. Chapters are a separate set of named or timed sections. Stream-level metadata belongs to an individual track; a language tag on an audio stream describes that stream, not necessarily the file or every audio track.
This distinction guides both the mapping plan and final checks. A file may have a global title but no chapters, or chapters but no language tag on its audio. None of these conditions establishes whether the other layers are complete.
FFmpeg’s documentation on demuxers, muxers, and stream copy describes demuxers as reading global properties such as metadata and chapters, while muxers write global properties and metadata. It also explains that stream copy transfers encoded packets without decoding or re-encoding, and can be used to change container-level metadata. These capabilities do not mean every field is compatible with every output container.
Make a field-by-field decision
For each item required by the receiving platform or your archive, record its source, destination, and action. Use three actions:
- Copy: retain a value from the input if it is accurate and applicable to the output.
- Set: write an intentional output value, such as a corrected title or an audio stream’s language tag.
- Omit: do not carry a value that is wrong, private, irrelevant, or unsupported by the delivery specification.
Do not treat “copy everything” as a substitute for deciding. A production title, for example, may be unsuitable as the public title. Likewise, do not label a track based on the video’s spoken language if that particular audio stream contains music only. Check the stream’s content before choosing its language value.
For a remux with explicit stream selection, FFmpeg documents -map for choosing input streams and -c copy for copying their encoded packets. Option placement matters: FFmpeg options generally apply to the next input or output file. Consult the FFmpeg stream-selection and option documentation for the exact mapping and stream-specifier syntax supported by your installed build. The mapping plan determines which streams enter the output; it is not a complete metadata policy by itself.
Work through the remux in order
- Inspect the input’s global metadata, chapter list, and each stream’s index, type, language tag, and other relevant tags. Record the values rather than relying on a player’s display name.
- Compare that inventory with the delivery specification. Mark each field or chapter set as copy, set, or omit, and note the input source and intended output target.
- Decide the output stream mapping separately. Confirm that every required video, audio, subtitle, or data stream is selected; automatic selection may choose only one stream of a type.
- Create a separately named output using stream copy where the selected streams and target container allow it. Apply explicit metadata or stream-tag options only for fields you intend to set; do not assume a global metadata option sets a stream-level language tag. The cited material supports stream-scoped option placement and stream specifiers, but does not show a language-tag option or guarantee its syntax across builds. Before using such a setting, verify the installed build’s official option documentation, record the exact command or tool method in the worksheet, and then inspect the output stream field.
- Re-read the completed output. Compare global fields, chapter count and titles or times, and each mapped stream’s language tag against the worksheet. Also confirm that the intended streams are present. For a language tag, record the exact stream-scoped setting method used and verify the resulting stream field; the cited FFmpeg material establishes stream specifiers and option scope, but does not provide a language-metadata command example.
- If a required value is missing or wrong, revise the mapping or metadata decision and create another candidate. Keep the source intact and validate the new candidate rather than inferring success from FFmpeg’s exit status.
Example acceptance worksheet
The values below are hypothetical. They illustrate the decision process, not a tested command or a guarantee about a particular container.
| Item | Input observation | Decision and output target | Output check |
|---|---|---|---|
| Global title | GenShort_v7 | Set file title to Glass Garden | Output global title reads Glass Garden |
| Global comment | Contains an internal review note | Omit from public output | No review note in global metadata |
| Chapters | 3 entries at 00:00, 00:18, 00:42 | Copy if supported by target container | Output has 3 entries with expected times and labels |
| Audio stream 0 | Narration; tag und | Set the stream language with a verified stream-scoped method supported by the installed build; record that method | Output audio stream 0 has the intended language tag and the worksheet records how it was set |
| Audio stream 1 | Music-only; no language specified | Omit a language claim unless delivery rules require one | Output does not falsely label the music as spoken dialogue |
Treat each output check as an observed result. If the chapter count matches but a title is missing, the chapter layer is only partly acceptable. If a language tag is correct on stream 0 but stream 1 was accidentally omitted, the overall delivery still fails its stream requirements.
When a field does not survive
First identify the layer and field that failed. If a stream is absent, revisit -map; metadata settings cannot restore an unmapped stream. If global metadata is wrong, check whether the intended output value was set for the output file and whether the target container supports it. If chapters are missing or altered, verify that they existed in the input, that the output format can represent them, and that the output actually contains the expected chapter entries.
If the target format cannot store a required item, choose a compatible container or deliver that information separately when the receiver permits it. Stream copy avoids decoding and re-encoding selected packets, but it cannot apply filters and may fail when the target container lacks required stream information. A correct language tag helps identify a track; it does not guarantee that every player will select that track or display the tag consistently.
Does `-c copy` preserve all metadata and chapters?
Do not assume that it does. Stream copy concerns encoded stream packets. Make separate decisions for global metadata and chapters, then inspect the output.
Is a language tag the same as making an audio track the default?
No. The language value describes the stream. It does not by itself establish default-track behavior or dictate how a player chooses among tracks.
Should I keep chapters if the short has no scene breaks?
Only if they serve the delivery or archive. If chapters are not required or are misleading, record an omit decision and verify that the output does not contain unintended entries.
Sources and next step
This guide interprets public documentation and uses explicitly hypothetical acceptance examples, not application tests. Controls and availability can change; verify the cited manual against your installed version.
For work created in Imagild, carry the input requirements and acceptance checklist into the film workspace. The external software controls discussed above belong to their respective products.