Bulk Export: Many posts at once — paid plans only. See pricing

TikTok Tools

TikTok Voice Comments — Free Download to CSV, Excel & JSON (2026)

By Nuttapong S., Product engineer, ZocialCommentSeptember 20269 min read
TikTok Voice Comments — Free Download to CSV, Excel & JSON (2026)

Get every comment from any TikTok — free

Free — first 100 comments of any post, no signup. CSV, Excel & JSON.

If you have started seeing a microphone next to the comment box, TikTok's voice comments have reached you. It is a small feature with an awkward consequence for anyone who works with comment data: for the first time, a meaningful share of a TikTok comment section is not text at all. This guide covers what a voice comment is, what happens to it when you download TikTok comments free to CSV, Excel or JSON, why those rows arrive blank, and how to keep a usable record of a comment section that has stopped being purely typed.

What TikTok's voice comments actually are

The mechanic is deliberately simple. A microphone icon appears at the edge of the comment field; press and hold it, speak, release, and the recording posts into the thread as a comment with a little waveform where the words would be. It moved through testing from around May 2026 and into a gradual wider rollout, as Techweez reported in September 2026. Gradual is the operative word: availability differs between accounts on the same app version in the same country, which is why half the comments on a thread about the feature are people saying they do not have it.

Voice comments join a category that already existed on TikTok and that most people never thought about — the comment that carries no text. Photo replies and sticker or GIF replies have been there for a while. Voice notes make the category big enough to notice, and big enough to break a workflow that assumes every comment is a string.

The blank row: what a voice comment looks like in a downloaded file

Here is the honest version, because it is better to know before you download than after.

A voice comment comes back as a complete, ordinary row. You get the commenter's display name and handle, their avatar link, the like count, the reply count, the timestamp, the comment id, and the id of the parent comment if it sits inside a thread. Everything that identifies who spoke and where and when survives intact. What is missing is text, because there is no text — nobody typed anything, and no tool can export characters that were never entered.

That gives you a useful trick and an honest limitation at the same time. The trick: sort by text ascending and every non-text comment in the post floats to the top of the sheet in one move. The limitation: the audio is not in the file and cannot be, so a voice note in a downloaded comment list is a pointer, not a transcript.

Photo and sticker replies land in the same blank-text pile but are distinguishable, because TikTok returns a media URL for them. The export carries image_url and image_type, where image_type reads photo for a user-uploaded picture and sticker for a TikTok sticker or animated GIF. So the sorting rule is: blank text with an image_url is a picture, blank text with no image_url is a voice note. One filter separates them.

One warning about those media links that catches people out. TikTok signs them, and the signature expires after roughly thirty days. The row is permanent; the picture behind it is not. If image replies matter to your research, fetch them within the first few weeks rather than treating a spreadsheet from March as an archive you can open in July.

Downloading a comment section that is no longer only text

The mechanics do not change because voice notes exist — you still paste a link and get a file — but two settings matter more than they used to.

  1. Copy the video link. Share, Copy link, or the browser address bar. Short links, full /video/ URLs and links with tracking parameters all resolve to the same video id.
  2. Paste it into the TikTok comment exporter. No login, no account connection, no browser extension required.
  3. Switch replies on. This is the setting that matters now. Replies are off by default because expanding them multiplies your row count, but voice notes and image replies concentrate inside reply threads rather than at the top level — they are a conversational move, not an opening statement. A top-level-only download of a chatty post will under-count them badly. The guide to exporting TikTok comment replies covers what changes when you do.
  4. Pick a format. CSV for anything, Excel if a human is going to filter it, JSON if a script or a model reads it next.
  5. Sort by text, then filter on image_url. Two clicks and your non-text comments are separated into pictures and voice notes.

The measurement problem nobody warns you about

If you count engagement, voice comments quietly change your denominators, and it is worth naming exactly how.

Sentiment scores drift optimistic. Almost every sentiment pipeline scores blank text as neutral. A post that provokes fifty furious voice notes and two hundred typed comments will read as calmer than it is, because the fifty loudest reactions score zero. If your dashboard says a launch went fine and your inbox says otherwise, count the blank rows before you trust the dashboard.

Keyword counts undercount. Searching a comment file for a product name, a competitor, or a giveaway keyword can only match typed characters. A voice note saying the keyword is invisible to the search and always will be.

Language detection has nothing to work with. The language column is derived from text. On a voice note it is empty, which will skew any per-country breakdown built on comment language toward whoever types.

The fix is not clever, it is arithmetic: report the blank-text share alongside every number you publish. "412 comments, 37 of them voice or image replies not included in the sentiment split" is an honest chart. A pie chart of 412 comments that silently treats 37 as neutral is not.

Giveaways: voice entries are the new argument

Giveaway hosts have a specific problem here, and it arrives about two hours after the draw.

If your rules say "comment your favourite colour to enter" and an entrant records a voice note saying their favourite colour, they have followed the rules as a human understands them and failed them as a spreadsheet understands them. Their row has no text, so no keyword filter will ever match it, and they will appear in your replies asking why they were excluded.

Two defensible ways out, and you have to pick one before the giveaway opens rather than after somebody complains:

  • Say entries must be typed. One line in the rules, and every entry is machine-checkable. This is the boring, correct choice for a large draw.
  • Accept voice entries and handle them by hand. Download with replies on, filter to blank-text rows with no image, listen to each one, and add them to the eligible list manually. Workable at a dozen; miserable at three hundred.

Either way, publish the eligible entry count with the winner. The mechanics of a draw you can defend are the same as they have always been — the TikTok comment picker walkthrough covers the deduplication and the timestamp cutoff, and the free comment picker runs the draw itself.

How we know this

We build and operate the exporter, so what we can speak to is the shape of what passes through it in aggregate, not any individual account's numbers.

The blank-text row is not new to us. Sticker and photo replies have been landing with empty text since long before voice notes, which is exactly why the export carries image_url and image_type as separate columns rather than trying to squeeze a picture into a text field — a decision made because users kept writing in about "empty" comments that were not empty on the page. Voice notes slot into the same handling for free, and the fact that they arrive with no media URL is what makes them separable.

The gap between the number on the video and the number of rows in a file is the single most common question we get, and it is worth stating plainly that it is not a truncated download. The badge on a TikTok video counts comments the platform has accepted; a download returns comments the comment list will currently serve. Deleted comments, comments from accounts that were later removed, and comments filtered by the creator's own keyword list sit in the first number and not the second. Voice comments count in the badge too, and they do produce a row — they just produce one you cannot read.

The column names above, the free cap of 100 comments per post, and the roughly thirty-day media link expiry are the live product values at the time of writing, not marketing approximations. On the platform's own rules for comments — which apply to voice notes exactly as they do to typed ones — TikTok's Community Guidelines are the primary source, and they are explicit that comments and audio are both in scope for moderation. For what you may then do with comment data about identifiable people, the GDPR overview and the California Attorney General's CCPA page are the two most people are actually bound by.

Columns worth knowing when comments stop being text

Every download carries author, username, avatar_url, text, likes, replies, created_at, language, is_pinned, id and reply_to_id, plus TikTok's own raw fields. Five earn their keep on a mixed-media comment section:

  • text — empty is now information, not an error. It is your non-text filter.
  • image_url and image_type — separates photo replies and stickers from voice notes, and gives you the picture while the link lives.
  • reply_to_id — empty for top-level, populated for replies. Voice notes cluster here, so this is how you find them rather than scrolling.
  • created_at — the only reliable deadline enforcement, and the fastest way to spot a burst of coordinated entries.
  • likes — a voice note with two hundred likes is worth the ninety seconds it takes to go and listen to it. This column is how you decide which blanks are worth opening.

That last one is the practical resolution of the whole problem. You cannot read voice comments in a spreadsheet, but you can rank them, and ranking turns "listen to three thousand comments" into "listen to the eleven that people liked".

Download TikTok comments free →

Related reading: how to export TikTok comments, how to analyze TikTok comments, saving pictures from TikTok comments.

Get every comment from any TikTok — free

Paste any TikTok video URL — every comment in CSV or JSON in seconds.