How to Transcribe a Full Livestream VOD (Up to 10 Hours)
A whole Twitch, Kick, or CHZZK broadcast is 6-10 hours of unedited audio. Here's how to turn one into a complete, timestamped transcript in a few minutes — and what it actually costs.
Clip transcription is a solved problem. Whole-broadcast transcription is the one that keeps breaking.
A streamer goes live for seven hours. The VOD expires in 14 days on Twitch, or gets buried on Kick. Somewhere in there is the segment you actually want — the sponsor read, the announcement, the take that got clipped out of context. Scrubbing a timeline to find it is miserable, and most transcription tools quietly refuse the job: they cap out at two or four hours, or they accept the file and hand back a transcript that stops a third of the way through.
SubGrab now handles livestream VODs up to 10 hours in a single pass. This post explains what changed, what it costs, and where the real limits still are.
Why Long VODs Break Most Tools
A livestream VOD is not a video file in the normal sense. It's a recording of a broadcast, stored as thousands of small HLS fragments, and — critically — most streaming platforms don't publish an audio-only version of it. A podcast gives you a 60 MB MP3. A seven-hour Kick VOD gives you a video rendition and nothing else, so anything that wants the audio has to pull video down to get at it.
That's why length caps exist. At full quality, a seven-hour broadcast is tens of gigabytes. Downloading that to extract speech is absurd, so tools either refuse long VODs outright or time out halfway and call it a failure.
The way through is to stop treating quality as the goal. Transcription only needs the audio track, and on these platforms every video rendition carries the *same* audio. So the correct move is to take the smallest rendition that still has sound — 160p, which nobody will ever watch — and throw the pixels away. The speech is bit-for-bit identical to what's in the 1080p60 version, at a fraction of the size.
What It Costs
Pricing is per hour of content, and it doesn't change for long videos:
| VOD length | Credits | Popular pack ($11.99 / 30 credits) |
|---|---|---|
| 1 hour | 1 | $0.40 |
| 3 hours | 3 | $1.20 |
| 6 hours | 6 | $2.40 |
| 8 hours | 8 | $3.20 |
| 10 hours | 10 | $4.00 |
Two things worth knowing before you paste a link:
- You see the cost before you're charged. A long VOD is quoted up front — the exact credit count appears next to a confirm button, and nothing is deducted until you click it. There is no way to accidentally spend 8 credits.
- Failures are refunded automatically. If the download or transcription fails for any reason, the credits go straight back to your balance. You can see every movement in your dashboard.
Because a new account starts with 2 free credits, a full 8-hour broadcast is not something you can transcribe for free — that's an honest limitation rather than a hidden one. Use the free credits on a clip or a shorter video first to see whether the output is what you need.
Platform by Platform
- [Kick](/kick-transcript) — Kick VODs are whole unedited broadcasts, routinely 6 to 8 hours. All supported up to 10 hours. Kick publishes no captions at all, so every transcript is AI-generated from the audio.
- [Twitch](/twitch-transcript) — Clips and VODs both work. Twitch VODs expire (14 days on most accounts, 60 for Partners), so transcribing is also an archiving move: the text outlives the video. See saving Twitch VOD transcripts before they expire.
- [CHZZK](/chzzk-transcript) and [SOOP](/soop-transcript) — Korea's two big streaming platforms, same 10-hour ceiling, with strong Korean-language accuracy. See the East Asia platform guide.
Do Timestamps Stay Accurate at Hour Nine?
This is the question that matters for long content, and it's where a lot of tools quietly fall apart. Timestamps drift when a transcript is assembled from chunks that each restart their clock at zero.
SubGrab splits long audio into 45-minute pieces, transcribes them, and then offsets each piece by its real position before merging. A line at 8:42:15 is genuinely at 8:42:15. That's what makes the transcript useful as an index — you search for "sponsor", get a timestamp, and jump straight there in the original VOD.
Use the Summary for Long VODs
For anything over about two hours, read the AI summary before the transcript. A ten-hour broadcast produces an enormous wall of text; the summary gives you a chapter breakdown with timestamps, so you can find the twenty minutes you care about and read only those.
One honest caveat: at the extreme end — past roughly seven and a half hours of dense speech — a transcript exceeds what a summarisation model can read in one pass. When that happens the summary says so explicitly rather than pretending it covered the whole thing. The transcript itself is always complete; it's only the summary that's bounded.
What Still Won't Work
- Uploaded files cap out around 50 minutes, not 10 hours. Your browser extracts the audio before sending it, and that extracted audio hits a 100 MB upload limit long before the duration limit matters. For long content, paste the URL instead of uploading.
- Active live streams — the broadcast has to have ended and the VOD published.
- Deleted or expired VODs. Twitch in particular deletes on a timer. Transcribe while it exists.
Try It
Paste a VOD link and you'll see the length, the credit cost, and a confirm button before anything is charged.