tracklog.

Prepare a big spreadsheet with AI

Most spreadsheets import as they are. A sheet kept for years often has its own system. The AI chat you already use can rewrite it for tracklog. Copy the prompt below and attach your sheet. Then check how the AI read it.

Do you need this?

Try the import first. It reads most sheets on its own. See Import a spreadsheet.

Use the prompt when your sheet has its own system, like one of these:

  • The status is only a colour. A green cell means live, a red one means rejected. A CSV file has no colours.
  • One sheet for each platform. A Pond5 sheet, an AudioJungle sheet. The same track is on several of them.
  • One row for each upload. Date, track, platform, status. A track takes many rows.
  • One row for each version. Main, 60s, 30s, Loop, each on its own row.
  • Notes inside titles. Sunrise (REJECT pond5), Lofi 3 (old name: Chill Beat).
  • Headings and totals between the tracks. — 2019 —, LOGOS, a total at the bottom.

What you need

  • The AI chat you work in: ChatGPT, Claude, Gemini or another. It needs to read a file you attach.
  • Your spreadsheet as an Excel file (.xlsx). In Google Sheets: File → Download → Microsoft Excel. An Excel file keeps the colours and every sheet. A CSV file keeps only one sheet and no colours.
  • Only your catalog in the file. Delete sheets with private things first: bank details, contacts, passwords.

The prompt

You are converting a musician's music catalog spreadsheet into one exact CSV format, so it can be imported into a desktop app called tracklog. I attach my spreadsheet. The target format is described below, with an example at the end of this message.

**Language.** Everything you show me is in the language of my messages — including the point template, the wording and the line «Here is how I read your sheet…» below: those are written in English here only as a pattern, **translate them**, never copy them word for word. Don't mix languages in one message. The CSV itself stays in the source's own language: never translate titles, genres, tags or notes.

## Read the file with code, not by eye

Open my file with code (Python: `openpyxl` for .xlsx, `csv` for .csv). Look at **every sheet, including hidden ones**, before you ask anything. For .xlsx also read what plain text loses:

- **cell fill colours** in every column — a colour is often the only place a status is written;
- hyperlinks behind cells, cell comments, merged cells, strike-through text;
- hidden sheets, hidden rows and columns;
- formulas: read each cell's value, not the formula text (`openpyxl` with `data_only=True`); a `HYPERLINK(...)` formula also gives the cell a link;
- each date cell's number format: a cell shown as «03.2024» or «Mar 2024» holds the 1st of the month, but only the month is meant.

Sheets that are not the catalog — bank details, contacts, passwords, invoices — are private. Don't read further into them and never repeat what's in them; just say you left them out.

Build the final CSV **with code from the file's data**, never by retyping rows: a catalog can have hundreds of tracks, and a retyped table loses rows and changes letters. Check the count before you hand it over.

If you can't open files or run code, say so in one line and ask me to paste the table as text, in parts. If my file is a .csv and I mention colours, tell me to send the .xlsx instead (in Google Sheets: File → Download → Microsoft Excel), because a .csv has no colours.

## How we work — two steps

**Step 1. Tell me how you read my file. No CSV yet, no mapping table.**

Study my file quietly. Decide everything obvious yourself and don't mention it: a column called "Envato Title" is the title, TRUE in a platform column is `Live`, `P5` is Pond5, a target column my file has no data for is simply left out.

For everything that changes the data AND can't be worked out from the file, don't ask me to pick from options. Say how you read it, and under it the other likely reading. I answer «ok», or I tell you in my own words what's different. I should never have to match numbers to letters.

There is no limit on the number of points — a big file can need many; don't fold two unrelated things into one point just to keep the list short. If there are more than about 10, send them in parts of up to 10, most important first; my answer covers that part, then send the next one. Build the CSV only after the last part. If nothing needs checking, say so in one line and ask me to reply «ok» to get the file.

Your whole reply is:
- one line on what you found: sheets, how the file is built (see "Shapes of spreadsheets"), how many tracks;
- one line: «Here is how I read your sheet. If it's all right, reply «ok». If not, tell me what's different in your own words.» (in my language);
- the points, each one laid out like this.

**Formatting — this is read in a chat window, make it pleasant to read:**
- Use normal chat formatting (markdown). **Never put the points in a code block** — no triple backticks, no monospace, no hard line breaks inside a sentence.
- Each point starts with its number and your reading in **bold**, on its own line: a plain statement of what you will do. Real values from the file go in it in «quotes», 2–3 of them, short (cut long ones with …).
- Under it, in *italics*, "In tracklog:" (translated) — what I will see after the import, and whether anything is lost.
- Under that, "Or:" (translated) — the other likely reading, with its own "In tracklog:" on the same line. Two other readings at most. Leave "Or:" out when there's no sensible other reading.
- One empty line between points. No horizontal rules, no tables, no headings inside the points.

The shape (the angle brackets are placeholders, not text to copy):

**1. <What you will do, as a plain statement, with «<value>», «<value>» from the file>**
*In tracklog: <what I will see after the import, and whether anything is lost>*
Or: <the other likely reading>. *In tracklog: <…>*

Two filled-in examples (shown in English; write yours in my language):

**2. The column "Name second" («i do», «jillin - vlogs», «Calm Music/ gentle») looks like working names and notes. I'll keep it as it is.**
*In tracklog: every track card gets a text box called "Name second" with exactly this text. Nothing is lost; you can rename or delete it later in Settings.*
Or: it's the track's name on a stock site, and you tell me which one. *In tracklog: shown on the track card next to that site as "the name there".*

**3. A green cell means the track is live there (34 cells, e.g. «Epic Rise» on Pond5).**
*In tracklog: a green dot for that platform.*
Or: green means "sent, waiting". *In tracklog: an amber Review dot.*

Rules for the points:
- Write for a musician who has **never opened tracklog**. They don't know what a "field", "target", "alias", "mapping" or "extra column" is — never use those words. Say what they will **see**.
- **Every reading has an "In tracklog:" line** that says, in one or two plain sentences, what the person will see in the app after the import and whether anything is lost. Take these only from "What each choice becomes in tracklog" below — never invent how the app behaves.
- In the rules below, «**A.** … *— recommended*» is your reading (the bold line), and «**B.**», «**C.**» are the other readings (the "Or:" line).
- **Group similar things into ONE point.** All platform abbreviations go into one point: a short list, one line each — **aj** → Envato / AudioJungle — and "Or:" says: tell me which ones are wrong, e.g. «zec is Zedge». List only guesses you are fairly sure of. A header you can't decode (`Met`, `Cruc`, `STr`) is not a guess: put it in the unclear-columns point below, so «ok» keeps it as it is instead of renaming it wrongly.
- Unclear columns that don't fit any target: one point, a short list of them (each with 2–3 real values). Your reading: each stays as its own text box on the track card. "Or:" leave them out.
- When you can't make a sensible guess at all (a mark `x` with no hint what it means), it's still a point: say what you'll do if I don't tell you (usually: leave it empty), and ask what it means.
- Don't make points of things you'll only report (duplicate titles, empty columns).
- **Always check for the kinds of mess below** — "Shapes of spreadsheets", "Colours" and "Clean titles" — and, if you find them, make ONE grouped point each, with 3–5 real examples from the file.

Then stop and wait.

**Step 2. Build the CSV** from my answer, then give me a short report (see the end). «ok» means every reading as you wrote it. Anything I wrote in my own words changes only the points it is about; the rest stays as you wrote it. If my answer is unclear about one point, ask about that point only, once.

## Shapes of spreadsheets

The target is always one row per **track**. Spreadsheets get there in different ways. Find out which shape mine is (a file can mix them) and say it in the first line of step 1.

1. **One row per track.** The usual shape. Rows become tracks as they are.
2. **The same columns on several sheets** — one sheet per year, per genre, per album. Put all sheets together. If a sheet's name says something the rows don't (a year, a genre), ask whether to use it: **A.** use the sheet name as the year / genre of its tracks *— recommended* (only when the rows have no date / genre of their own).
3. **One sheet per platform** — a sheet "Pond5", a sheet "AudioJungle", each listing what is uploaded there, often under that site's own title. One track appears on several sheets. Join them into one row per track: by a shared code or ID (ISRC, a catalog number, a "master track" column) first, then by exactly the same title. The sheet's status goes to that platform's column; a title that differs from the main one goes to `Name on <Platform>`. Pairs you are not sure about (similar but not equal titles, no shared code) — make one point listing them, each pair with your guess: «Quiet Harbor» = «QUIET HARBOUR», same track; «Tech Pulse» and «TECH PULSE 2», different tracks. "Or:" tell me which guesses are wrong, e.g. «Tech Pulse 2 is the same track».
4. **One row per upload** — a log: date, track, platform, status, maybe a link. Join the rows into one row per track. When a track has several rows for the same platform, the latest by date wins, not the lowest row (no dates — the lowest row wins). The dates of a log are upload dates, not the day the track was written: don't put them in `Date`. Make it a point, with an example of a track whose status changed.
5. **One row per version** — "Main", "60s", "30s", "Loop", "Stems" rows for one piece of music. Often only the first row has the title and the rows under it are blank: a blank title means "the same track as above". One track; the version names go to `Versions`, their lengths to `Lengths` in the same order. A stems row is not a version: it gives `Stems` = `yes`. The track's status on a platform is the status of its main (full) version; no main version — the best status any version has. If versions differ on a platform, write it in `Notes`, like «Pond5: only 30s is live». Ask once, with examples: **A.** these rows are versions of one track *— recommended* / **B.** each row is its own track.
6. **Sheets and rows that are not tracks** — totals, statistics, a colour legend, a to-do list, a hidden sheet of old junk. Leave them out. Hidden sheets: ask, naming each one and what's in it.

In shapes 3, 4 and 5, joining rows is the point. Everywhere else, never join rows yourself (see Hard rules).

## Colours

People often mark a status only with a cell's fill colour: green = live, yellow = pending, red = rejected, or their own system. A colour you skip is a status lost.

- Check fill colours in every column, especially platform columns and the title column.
- If colours vary in a way that looks meaningful, find out what each one means. **Never guess a colour silently.**
- If the file has a legend (a small table "green = approved"), use it, and still show it as one point.
- No legend: make ONE point, listing each colour by a plain name (green, light yellow, red…), how many cells have it, and 2 example tracks with the column; for each colour, your best reading among the statuses below or "means nothing", and ask me to correct any that are wrong.
- A colour and the cell's value that disagree (a red cell saying «live», a date in a cell with no colour while the legend says «no colour = not uploaded») → make it a point: which one you'll trust, with examples.
- A colour over a whole row (not one cell) usually means something about the track (finished, exclusive, sold). Ask what it means; a meaning that fits no target column becomes its own text box (for example a "Colour" box holding "exclusive").

## What each choice becomes in tracklog

Use these, and only these, for the "In tracklog:" lines. What the person sees after import:

| Where the data goes | What they see in tracklog |
|---|---|
| `Title` | The track's name in the catalog table and at the top of its card. |
| `#` | The track's catalog number, shown next to its name. |
| `Stage` = In progress | The track goes to the "In progress" section instead of the catalog, and gets a catalog number only when marked ready. |
| `Date`, `BPM` | Shown on the track card; the catalog can be sorted by them. |
| `Genre`, `Subgenre`, `Mood`, `Use case` | Chips on the track card, picked from a list; the catalog can be filtered by them. |
| `Tags` | Keyword chips on the track card; found by search. |
| `Description` | The buyer-facing text box on the track card. |
| `Notes` | The track's private Notes box on its card. |
| `ISRC`, `ISWC` | The Codes section of the track card. |
| `Versions` + `Lengths` | The track's list of versions with their lengths on the card. |
| a platform column | A column of status dots in the catalog table for that platform: Live, To upload, Review, Reject — or empty. |
| `Name on <Platform>` | On the track card, next to that platform: "the name there", when it differs from the title. |
| `Release`, `UPC` | The release (album / EP / pack) the track belongs to, with its code. |
| kept as its own column at the end | A new text box with the column's name on **every** track card, holding exactly this text. Nothing is lost. It can be renamed, turned off or deleted later in Settings. |
| kept at the end, only yes / no or TRUE / FALSE in it, and not a platform | A tick box with the column's name on every track card. |
| left out | Not imported at all. It stays only in the spreadsheet. |
| a row left out (heading, placeholder) | Not in the catalog. |

## The target format

One row = one track. First row is the header. Columns in this order; leave out a column entirely only if my file has no data for it at all (Title is always required):

| Column | What goes in it | Format |
|---|---|---|
| `#` | Catalog number | As in my file (`24`, `A-017`). Empty for unfinished tracks. Never invent numbers. A column that only counts rows (1, 2, 3… or a formula) is not a catalog number: leave it out and say so in the report. |
| `Title` | Track title | Required. Exactly as in my file. Several title columns: the main one is the one with no platform in its header («Title», «Name»); the others go to `Name on <Platform>`. No such column, or the file itself says which name is main — ask. |
| `Stage` | Finished or not | `Catalog` (finished, released, submitted or ready) or `In progress` (work in progress, draft, idea, mixing). Only when my file says which tracks are unfinished; otherwise leave the column out (tracklog puts everything in the catalog). |
| `Date` | When the track was written or finished | `YYYY-MM-DD`, `YYYY-MM` if only the month is known, or `YYYY` if only the year. Never invent a day or a month. A range like «2023-03-01 → 2023-04-10» → its start. Only a release date in the file → use it and say so in the report. |
| `BPM` | Tempo | A number. |
| `Genre` | Genre | Several values separated by `; ` |
| `Subgenre` | Subgenre | Several separated by `; ` |
| `Mood` | Mood | Several separated by `; ` |
| `Use case` | What the track is for (vlog, corporate, trailer…) | Several separated by `; ` |
| `Tags` | Keywords | Separated by `; ` |
| `Description` | Description written for buyers | Free text |
| `Notes` | Private notes | Free text |
| `ISRC` | Recording code | As written, e.g. `QZABC2400001` |
| `ISWC` | Composition code | As written, e.g. `T-123.456.789-0` |
| `Versions` | Names of the track's versions/cuts | Separated by `; `, e.g. `Main; Loop; 60s Cut` |
| `Lengths` | Durations of those versions | Separated by `, `, **same order as Versions**, e.g. `2:04, 0:26, 1:00`. A single length with no version names is fine on its own. A length stored as a time of day (`00:37:00`) is often minutes and seconds typed as hours and minutes: if the hours reading makes tracks hours long, it's `0:37`. Ask once, with examples. |
| `Stems` | Are stems included | `yes`, or empty |
| *one column per platform* | This track's status on that platform | One of: `Live`, `To upload`, `Review`, `Reject`, or empty |
| `Name on <Platform>` | The track's title as that platform knows it, when different from Title | Text. One column per platform, e.g. `Name on Pond5`. |
| `Release` | Album / EP / single / pack the track belongs to | Several separated by ` \| ` |
| `UPC` | The release's UPC | As written. A UPC with no release name → put it in `Notes` as «UPC 195999000011». |

### Platform columns — the most important part

Whatever the source shape, the target is always **one column per platform, one status per cell**:

- A cell listing platforms ("Uploaded: Pond5, MotionArray") → a `Live` in each of those platform columns for that track.
- A cell with platform + state ("BMI — pending") → `Review` in the BMI column.
- A checkbox / TRUE / ✓ / "yes" column per platform → `Live`; FALSE / empty → empty.
- A date in a platform cell (the day it went up) → `Live`, unless its colour or another mark says otherwise. Ask once what to do with these dates: **A.** keep them in `Notes`, like «Pond5: 2024-03-15» *— recommended* / **B.** leave them out.
- One status column for the whole row next to a cell listing platforms (Stores: «Spotify, Apple Music», Status: «Processing») → that status in each listed platform's column.
- A status written in notes or another text column («rejected by AJ», «PB said no») → also that status in the platform's column; the note stays in `Notes`. Ask once, listing them.
- A name in a platform's own name column (`Pond5 title`) means the track is on that platform: `Live`, unless a status for that platform says otherwise.
- One "names" column that covers several platforms (e.g. a header "MA 100Audio Pond5 Title") → copy the same name into `Name on MotionArray`, `Name on 100audio` and `Name on Pond5`. If a platform also has its own name column and the two differ, its own column wins.
- A link to the track's page on a site (in a name column, a status column, or behind a cell) is not a name. It still tells that the track is on that site: → `Live` there, unless something says otherwise (a later row, an unticked checkbox for that site, a note «deleted», «removed»). Ask once what to do with the links: **A.** move them into Notes, one per line, like «Pond5: https://…» *— recommended* / **B.** leave them out.

Status meanings:
- `Live` — the track is published / accepted / approved / uploaded / registered and live there.
- `To upload` — planned, not uploaded yet.
- `Review` — sent / submitted / processing / scheduled / waiting for moderation.
- `Reject` — rejected / declined by that platform.
- empty — nothing known, or the track isn't meant for that platform.

A word that fits none of these — `removed`, `taken down`, `deleted`, `expired`, `sold exclusively` — ask once, listing them. Recommend: empty in that platform's column, and the word with its date in `Notes`, like «Pond5: removed by me 2024-05». A mark you can't read (`x`, `?`, `-` in a platform column) — ask once what each means, with examples; never guess.

Name platforms like this when it's clearly the same site (use my file's own name for anything not on this list):
`Envato / AudioJungle`, `Envato Elements`, `Pond5`, `MotionArray`, `Artlist`, `Epidemic Sound`, `Storyblocks`, `Soundstripe`, `Uppbeat`, `PremiumBeat`, `Shutterstock Music`, `Musicbed`, `Music Vine`, `Melodie`, `audiocrowd`, `100audio`, `vFine`, `Pixabay`, `Jamendo`, `Free Music Archive`, `SoundCloud`, `YouTube Audio Library`, `DistroKid`, `YouTube Content ID`, `BMI`, `ASCAP`, `PRS`, `Bandcamp`.
Common short forms: MA = MotionArray, P5 = Pond5, AJ = Envato / AudioJungle, 100 = 100audio. A short form that could be two sites (PB = PremiumBeat or Pixabay? EP = Envato Elements or an EP release?) — ask.

"The same title" anywhere in this prompt ignores capital letters and extra spaces: «GOLDEN HOUR DRIVE» and «Golden Hour Drive» are the same title. Keep the title as written in the main sheet or column.

## Clean titles

A title cell often carries more than the title. The Title column must hold **only the track's name**. Look for these and fix them:

1. **Rows that are not tracks.** A row with a "title" and nothing else, where the title reads as a section heading — a label or brand in capitals (`VICATE`, `ATOMICA`, `ARTLIST IO`), a category (`Логотипы`, `Logos`, `SFX`), a year divider (`— 2019 —`), or a placeholder (`хз`, `???`, `test`). Leave these rows out. Ask once, listing them all: **A.** these are section headings, leave them out *— recommended* / **B.** keep them as tracks. A heading gives something to the tracks under it: a year → their `Date` if they have none; a category (`LOGOS`, `SFX`) → ask where to keep it: **A.** a text box "Section" on each track *— recommended* / **B.** add it to Genre / **C.** leave it out.
2. **Notes inside the title.** Brackets or tails that are about the file, not the name: `(REJECT)`, `(нет проекта)`, `(broken flp)`, `- New -`, `(old name: …)`, `X или Y` / `X or Y`. Title keeps the name; the note goes to `Notes` word for word. A tail that names a different mix or arrangement — `- new mix`, `(piano)`, `(Instrumental)`, `(Remix)`, `(feat. …)`, `(Extended)` — is part of the name: keep it, or two different tracks get one name. Unsure — ask, with examples. If the note is a platform status (`REJECT pond5`), put it in that platform's column; if it doesn't say which platform, ask. Ask once, with examples: **A.** move the notes to Notes *— recommended* / **B.** leave titles as they are.
3. **A title that is only a link** (`https://…/item/fashion-hip-hop/123`). Ask once, with examples: **A.** take the name from the link: the words of its last part, as they are, first letters capital («Fashion Hip-Hop»), and the link goes to `Notes` *— recommended* / **B.** keep the link as the title.
4. **Statuses inside the title.** `Rise Above (AO-001560) — website: available | Spotify: available`: the name is `Rise Above`; the code goes to Notes; each `platform: available` becomes `Live` in that platform's column, `NOT` stays empty. Ask once, with examples.

Also, without asking:
- A line break inside any cell → one space. (In `Notes` you may put several notes on separate lines.)
- Spaces at the start or end of a cell → removed.
- A cell holding only `-`, `—`, `--`, `n/a`, `?`, `No`, or the column's own header repeated (`Style` under «Style») in a non-platform column → empty.
- A tag cell that starts with a label (`Main instruments: sub booms; …`) → drop the label, keep the tags.
- A value copied from a section heading into another column of that heading's row → empty (the row is left out anyway).
- Text that is struck through → treat as removed, and mention it in the report.

## Hard rules

1. **Never invent data.** No made-up tags, descriptions, genres, codes, dates or numbers. Empty in my file = empty in the CSV.
2. **Keep values exactly as written** — spelling, capitalisation, language. Only change the *shape* (which column, which separator, the four status words, the date format).
3. **Keep every track.** Skip only what is clearly not a track (section dividers, totals, empty rows, legends). If a divider row gives a year to the tracks under it, and those tracks have no date, use that year as their `Date`.
4. **A column that fits nowhere** — sales, earnings, links to files, your own marks — keep it as an extra column at the end with its original header. If that header is the same as a target column's (`tags`, `UPC`, `Lengths`), add « (my sheet)» to it, so tracklog doesn't read it as that column. A header that only says which platform it is about (`Item ID` on the AudioJungle sheet) gets the platform: `AudioJungle Item ID`. The app will offer to make it a text box. Don't drop information silently.
5. **Duplicate titles** in a one-row-per-track sheet: keep every row, never merge them yourself — tracklog decides on import (rows that contradict each other become separate tracks, the rest merge). List them in the report. Joining rows is only for shapes 3, 4 and 5 above, after my answer.
6. **CSV rules:** UTF-8, comma as the delimiter, a value that contains a comma, quote or line break goes in double quotes (`"2:04, 0:26"`), a double quote inside a value is doubled (`""`).

## Output of step 2

1. The CSV as a downloadable file named `tracklog-import.csv` (if you can't make files, one code block containing only the CSV).
2. A short report, in my language:
   - tracks in my file, rows in the CSV, and which rows or sheets you left out and why;
   - rows you joined into one track (shapes 3–5), with two or three examples;
   - status words and colours you converted, with counts ("ok → Live: 212", "green → Live: 40");
   - cells you couldn't interpret and left empty;
   - extra columns you kept at the end;
   - duplicate titles, including titles that became equal after cleaning;
   - totals in my file that don't match your count (a "Stats" sheet says 11, you found 10).
3. One last line: "Now open tracklog → Import and choose tracklog-import.csv." (in my language).

## Example of the target CSV

```
#,Title,Stage,Date,BPM,Genre,Subgenre,Mood,Use case,Tags,Description,Notes,ISRC,ISWC,Versions,Lengths,Stems,Pond5,MotionArray,Envato Elements,100audio,BMI,Name on Pond5,Name on MotionArray,Release,UPC
1,Extreme Sports,Catalog,2024-03-15,140,Rock,Hard Rock,Epic / Powerful; Upbeat / Energetic,Sports; Action,guitar; drums; energetic; extreme,Driving rock with heavy guitars and big drums.,,QZABC2400001,T-123.456.789-0,Main; Loop; 60s Cut,"2:04, 0:26, 1:00",yes,Live,Live,Review,To upload,Live,Extreme Sports Rock,Adrenaline Rush,,
2,Travel Vlog,Catalog,2024-06,110,Pop,Indie Pop,Happy / Cheerful; Bright,Vlog; Travel,ukulele; whistle; summer,,Re-upload to Envato after new master,QZABC2400002,,Main; Short,"2:31, 0:45",,Live,Reject,,Live,,Travel Vlog Ukulele,,Summer EP,195999000011
3,Urban Vlog Hip Hop,Catalog,2023,90,Hip-Hop / Rap,Lofi / Chill Beats,Chill / Mellow,Vlog; Lifestyle,lofi; beat; urban,Laid-back hip hop beat.,,,,Main,2:15,yes,Live,Live,Live,,Live,,,Summer EP,195999000011
4,Corporate Inspire,Catalog,2022-11-02,120,Corporate,,Inspiring / Uplifting,Presentation; Business,piano; strings; motivational,,,,,Main; Loop,"2:48, 0:30",,To upload,To upload,,,,,,,
5,Night Drive,In progress,2025,100,Electronic,Synthwave,Calm,,synth; retro; night,,Needs final mix,,,,,,,,,,,,,,
```
Download it as a file

Step by step

  1. Press Copy the prompt above.
  2. Open a new chat. Paste the prompt. Attach your .xlsx file. Send.
  3. Write in your own language. It asks and answers in the language you write in.
  4. It tells you how it read your sheet, point by point. Under each point it gives another way to read it. If everything is right, reply ok. If not, write what is different in your own words, like "yellow means rejected".
  5. It gives you a file named tracklog-import.csv and a short report. The report says which rows it left out and why. Read it.
  6. In tracklog, open Settings → Import & export and click Import tracks…. Choose tracklog-import.csv.
  7. Look at the import window. It shows what each column becomes. Click Import.

Nothing is saved until you press Import. Your catalog is backed up before the import. Undoing an import removes only what that import added. See Undo.

Your data

Your spreadsheet goes to the company that runs your AI chat. tracklog does not see it. If you don't want your sheet used to train their models, turn that off in the chat's settings first. In ChatGPT: Settings → Data controls → Improve the model for everyone.

It says it can't open my file.

Try another AI chat. Or paste the table as text when it asks, in parts of about 100 rows.

Some tracks are missing after the import.

Read the report first. It names every row it left out: headings, totals, empty rows. If a real track is missing, tell it in the same chat: "Track X is missing, put it back". Then download the new file.

My colours came out wrong.

Check that you attached the .xlsx file, not a CSV. Then tell it in the same chat what each colour means, for example "green is live, yellow is pending". It makes the file again.

I don't understand one of its points.

Ask it in the same chat, in your own words. Or reply ok. After the import you can fix any track with the paint brush, or undo the whole import and try again.

Can I import again after fixing the sheet?

Yes. Undo the first import, then import the new file. Or import it over the first one and choose Fill in the blanks: it adds only what is missing.