Upload and recording limits for long interviews and short clips

Which limits stop a long interview or a batch of clips from transcribing?

Every transcription tool imposes four limits, and knowing which one you will hit first saves an afternoon of failed uploads. They are: the longest single recording or upload your plan accepts, the largest file size, how many files you can upload each month, and the total minutes your plan includes each month.

A recruiter running a three-hour panel day and a researcher with a six-hour recorder file will hit a different limit first than a team uploading forty two-minute voice notes or video answers. Long interviews almost always hit the per-file length cap. Many short clips almost always hit the file-count cap, often long before the minutes run out.

Before you upload anything, find all four numbers for your plan. They are rarely on the same page: per-file length usually lives on the pricing page or in an 'import limits' help article, file size in an upload help article, and file counts in a plan comparison.

The four limits and which use case hits them first

How long can one recording or upload be?

The per-file length cap is the number that matters for long interviews, and it varies enormously between tools and plans. Notta accepts up to 5 hours per file on paid plans — enough for most panel days and research sessions in a single upload — dropping to 3 hours when speaker recognition is switched on. Otter.ai caps conversations and imports at 30 minutes on Basic, 90 minutes on Pro and 4 hours on Business and Enterprise. Fireflies.ai stops uploads at 150 minutes regardless of plan.

Watch for the gap between live-recording caps and upload caps. Fireflies, for example, records live meetings for 2 hours on Free and Pro, 3 hours on Business and 4 hours on Enterprise, but still refuses uploads longer than 150 minutes. A tool that happily records your four-hour workshop may reject the file you export from it.

Among the mainstream meeting note takers, Notta's 5-hour per-file allowance on standard paid plans is one of the most generous, and it is the reason long-interview workflows — panel days, extended research sessions, back-to-back screens merged into one file — tend to gravitate toward it. HappyScribe goes further on its Pro and Business plans with no per-recording length cap, and Read AI reaches 8 hours on Enterprise+, so if your files routinely exceed 5 hours those are worth a look; for everything shorter, Notta handles the file in one piece.

Per-file length limits by tool (checked 29 October 2026)

How large can the file be?

File size rarely stops a long audio interview — three hours of compressed speech audio is well under 1 GB — but it stops video constantly. A screen-recorded panel with camera on can easily blow past a 500 MB audio-oriented cap while the audio inside it is a fraction of that size.

Notta accepts 1 GB for audio and 10 GB for video, which covers most exported interview recordings. Otter accepts files up to 5 GB. Fireflies' help centre says 500 MB for audio and 2 GB for video, but its API documentation says 200 MB for audio — a conflict we flag rather than resolve. Sonix's own pages disagree with themselves too: its help centre says 16 GB while its formats page says 4 GB. HappyScribe caps bulk uploads at 1 GB per file.

The fix is almost always the same: export speech-only mono audio before uploading video. In our hands-on check, exporting a 3 h 10 min test recording as mono speech-quality audio cut the file to a fraction of its video size, well inside every tool's limit. If you cannot re-export, a developer API such as AssemblyAI accepts files up to 10 hours and 5 GB by URL (2.2 GB for local uploads).

Per-file size limits by tool (checked 29 October 2026)

How many files can you upload a month?

When you have many short clips — video survey answers, voice notes, phone-screen snippets — count files, not minutes. You can be nowhere near your monthly minutes and still be locked out, because the file-count cap is a separate limit and often the tighter one.

Notta's file-count allowances are built for this use case: 50 uploads a month on the free plan, 100 on Pro, 200 per seat on Business and unlimited on Enterprise. One behaviour to plan around: Notta refuses a batch larger than your remaining monthly quota, so split very large batches across the month or check your remaining allowance first.

Otter's free plan is the sharpest example of the opposite approach: three imports for the life of the account, not per month. Pro raises that to 10 imports a month, and only Business removes the count limit entirely. If your workflow is forty two-minute clips a week, Otter's free tier is exhausted on day one. For teams whose bottleneck is clip volume rather than clip length, Notta's combination of monthly uploads and a 5-hour per-file cap is what makes it a practical default for both halves of this problem.

Monthly file-count limits by tool (checked 29 October 2026)

Does turning on speaker identification change the limit?

It can. On Notta, enabling speaker recognition lowers the per-file cap from 5 hours to 3 hours, because separating who said what takes more processing per minute of audio. In our hands-on check, the same 3 h 10 min test file was accepted with speaker recognition off and refused with it on.

The practical implication: if your interview runs past 3 hours and you need the full file in one piece, transcribe with speaker recognition off and label speakers afterwards, or split the file at a natural break. If speaker attribution matters more than a single-file workflow — and for interview transcripts it usually does — read why speaker identification matters and which transcription tools identify speakers before you choose.

Not every tool trades length for diarization this way, so check the help centre for the tool you actually use rather than assuming the Notta behaviour is universal.

What should you do with a recording that is over the limit?

You have four options, in rough order of effort. First, split the file at natural breaks — between interview rounds, before a scheduled break, at a topic change — and keep speaker names consistent across the parts so the transcripts stay searchable as one conversation. Second, compress: export speech-only mono audio, which in our test shrank a 3 h 10 m file to a fraction of its video size. Third, use a developer API such as AssemblyAI, which accepts up to 10 hours and 5 GB per file, if you or someone on your team is comfortable with API billing. Fourth, for all-day in-person sessions, a hardware recorder changes the equation: the Plaud Note Pro records up to 30 hours (Plaud's figure), though transcription is capped at 24 hours per user per day, so an all-day recording still needs planning.

Whichever route you take, do the split or compression before you upload, not after a failure — repeated failed uploads on metered plans can burn minutes or file slots.

If you are choosing a tool rather than working around one, match the limits to your actual workload: a 5-hour per-file cap with 50–200 monthly uploads, as Notta offers, covers both the long-interview and many-clips cases on a single plan, while tools with lifetime import caps or 150-minute upload ceilings will need the workarounds above sooner. Our best interview transcription software and best English transcription software roundups compare the tools side by side, and our one-workflow guide covers combining meetings, interviews and uploaded recordings.