Best Discord Transcription Bots for Voice Calls
We invited 28 Discord transcription, recorder, caption, and meeting-note bots into real servers and, where the bot let us, ran a live voice call. Here is the feature-gap matrix, the graded transcript accuracy, and which bots are free, paid, or dead — as of July 2026.
Direct Answer
The best Discord transcription bot depends on the job. For multi-track podcast audio, Craig is the incumbent. For free live accessibility captions, Scripty or Scriber. For AI meeting notes posted in your server — a recap with decisions and action items, plus the ability to ask past calls — the field narrows fast: most bots paywall auto-join, memory, or retention, and several are offline or dead. In our July 2026 hands-on test, Discap was the only bot that combined free auto-join, an announced-and-opt-out consent flow, a free AI recap, and free ask-your-calls memory in one product. No other bot's free tier did all four.
Key Takeaways
- Do not rank Discord speech-to-text tools as if they all solve the same job. Recorders, live-caption bots, and AI meeting-note bots produce very different final artifacts.
- Most convenience is paywalled elsewhere: Craig, DiscMeet, NoteCat, NotesBot, and iTranslator each charge for auto-join; NoteCat charges for its
/ask; iTranslator gates all voice behind €24.99/mo. Discap keeps auto-join, recaps, and ask-your-calls free. - In our hands-on standardized voice test, Scripty mangled proper nouns and hallucinated caption lines from room silence, and Craig surfaced none of the facts at all — its free tier has no transcript, just an audio file behind a 7-day link.
- Consent is uneven. Craig flags recording with a nickname; Discap announces recording and honours per-user opt-out; DiscMeet, Scripty, and iTranslator announce nothing by default.
- The category is a graveyard with a few live exceptions. Several listed "bots" (Textional Voice, Speech To Text Bot, MeetMind, Spext, Ablebot, Audio recorder) are offline, dead, or uninvitable, and SeaVoice was offline during our test.
How we tested
Over July 2026 we invited 28 competitor and adjacent bots into freshly created Discord servers through a real account, captured every OAuth screen, welcome message, and command reply verbatim, and — for the bots that could actually be brought into a voice channel — played one standardized two-speaker "stand-up" dialogue into the call. The script was seeded with nine known facts (a 12% signups figure, 84 new members, a webhook retry-loop bug to fix by Friday, a tournament launching Saturday 7 PM CET, an overlay toned down to two colors described as "kaleidoscope", a meet-up in Reykjavik in October, a sponsor call moved to Tuesday, and a deck due before Monday) so transcripts could be graded fact-by-fact instead of by vibe.
The honest limitation: it was a single standardized two-speaker call played through one Mac microphone, so both voices reached Discord as one speaker. That means speaker separation (diarization) was never testable, and we omit any speaker-attribution grade. Treat every accuracy figure below as "in our July 2026 hands-on test" — a fact-recovery count from one scripted call, not an absolute score or a percentage you should extrapolate. Competitor features and prices are stated "as of July 2026" and change often; reverify before you buy. Bots that got a live, gradeable voice run were Craig, DiscMeet, Scripty, and Scriber; NoteCat's live leg was blocked by a microphone limitation in our test rig and NotesBot's by a tooling outage, so those two are graded on setup and output shape only.
Feature comparison: what's free, paid, or missing
This is the table most listing pages won't give you, because it takes a real invite and a real call to fill in. Legend: ✓ = free and built-in · Paid = paywalled · Partial = limited or conditional · ✗ = not offered. Verified hands-on, July 2026.
| Bot | Auto-join | Announce / consent | Live captions | AI recap | Action items | Ask your calls | MCP | Exports | Opt-out | Free vs paid |
|---|---|---|---|---|---|---|---|---|---|---|
| Discap | ✓ | ✓ | ✗ | ✓ | ✓ | ✓ | ✗ | Partial | ✓ | Free |
| Craig | Paid | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ | ✓ | Partial | Partial |
| DiscMeet | Paid | ✗ | ✗ | ✓ | ✓ | Paid | Paid | ✓ | Partial | Partial |
| NoteCat | Paid | Partial | ✓ | ✓ | ✓ | Paid | ✗ | ✓ | Partial | Partial |
| NotesBot | Paid | Partial | ✗ | ✓ | ✓ | ✗ | ✗ | ✓ | Partial | Paid |
| Scriptly | ✓ | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | Partial | ✗ | Partial |
| Scripty | ✓ | ✗ | ✓ | ✗ | ✗ | ✗ | ✗ | Paid | ✓ | Partial |
| Scriber | ✗ | Partial | ✓ | ✗ | ✗ | ✗ | ✗ | ✗ | ✗ | Free |
| iTranslator | Paid | ✗ | Paid | ✗ | ✗ | ✗ | ✗ | Partial | ✗ | Paid |
| SeaVoice | ✗ | Partial | ✓ | ✗ | ✗ | ✗ | ✗ | ✓ | Partial | Free |
Reading the matrix: Discap is the only row that combines free auto-join, a clean announce-and-opt-out consent flow, a free AI recap, and free ask-your-calls memory. DiscMeet and NoteCat also post recaps, but each paywalls auto-join and the memory layer (DiscMeet's MCP, NoteCat's /ask). Craig gives you multi-track audio but ships no recap and no memory. Scripty, Scriber, and SeaVoice put captions in the channel in the moment, with nothing to ask afterward. "Ask your calls" here means putting a question to your past meetings and getting an answer back — Discap does this over your meeting history; it is not full-text search, and Discap does not expose an MCP endpoint today.
What transcription actually gets wrong on Discord
Two proper nouns — "Reykjavik" and "kaleidoscope" — were our tell for real speech-recognition quality, because they break weaker models. Here is what we watched go wrong in the July 2026 test.
| Where it broke | What happened |
|---|---|
| Proper nouns and silence (Scripty) | "Reykjavik" came out "Rakyavik" and "kaleidoscope" came out "lighter collider-stove", and about ten caption lines were hallucinated from room silence after the dialogue ended. |
| No transcript at all (Craig) | On the free tier there is nothing to read: the words sit inside a lossless audio file behind a 7-day link, and transcription is a Patreon Tier 3 perk that lives on the download page, never in Discord. |
| Guessed speaker owners (single-mic input) | With both scripted voices arriving as one Discord speaker, a bot that infers action-item owners from context assigned them confidently and wrong. Attribution that keys each speaker to their own Discord audio stream — how Discap captures — is what gets owners right on real calls. |
The load-bearing caveat: with both voices on a single microphone, none of this tests speaker separation, so we omit any diarization score. The lesson that carries into real calls is the attribution one — a bot that guesses owners from context will misattribute lines, while one that keys each label to a Discord user's own audio stream will not. (NoteCat and NotesBot are not shown here: their live legs were blocked in our rig by a headless-microphone limitation and a tooling outage, not by anything the products did.)
Verified status and pricing (July 2026)
Server counts come from each bot's own OAuth screen or in-product /info at the time of testing; treat them as rough, since Discord's "used in N servers" badge counts installs, not living usage, and several high-count listings are dead. Pricing is what we saw in-product in July 2026.
| Bot | Servers (verified) | Status | Transcription pricing |
|---|---|---|---|
| Craig | ~565,000 | Alive — the incumbent recorder | Free records multi-track audio; transcription is Patreon Tier 3; download links expire after 7 days. |
| iTranslator | ~119,000 | Alive — translation-first | All voice features behind Pro at €24.99/mo (Premium €8.32/mo); even /voice help refuses on free. |
| Scriptly | ~94,500 | Alive — quietly operated | Free caption tier; six paid tiers from $4.99 up to a new $49.99 "Max". |
| SeaVoice | ~15,300 | Alive but offline during our test ("application did not respond") | Free; Premium has been "coming soon" for years. |
| Scripty | ~9,000 | Alive — accessibility captions | Free live captions (5 users, 3-hour sessions); audio recording is paid. |
| DiscMeet | ~2,890 | Alive, growing | 2 free meetings/mo (up to 2h); Pro $9.99/mo, Pro+ $19.99/mo unlock auto-join, MP3, MCP. |
| NoteCat | ~2,850 | Alive, growing (by the sesh team) | Free unlimited recording + transcript + summary, artifacts expire after 7 days; Premium $7.99/mo unlocks /ask, auto-join, and no-expiry. |
| NotesBot | ~2,480 | Alive | One-time 30-minute trial, then usage tiers from $3/mo (5h) to $40/mo (100h). |
| Scriber | ~180 | Alive but tiny (5+ years old) | Free; no summary layer, and nothing is stored after the call. |
The graveyard. Not every "top" listing is a living product. In our sweep, Textional Voice, Speech To Text Bot, and MeetMind were offline; Spext, Ablebot, and Audio recorder are stuck at Discord's unverified 100-server cap and can't be invited (Audio recorder's listing now just advertises the developer's new app); and SeaVoice — despite ~15,300 installs — did not respond during testing. A high install count is not a sign of life. If you are choosing a bot for a real community, invite it into a throwaway server and run one call before you rely on it.
Best for / not best for, bot by bot
These tools do different jobs. The mistake is assuming they all do the same one.
| Bot | Best for | Not the best fit when (July 2026) |
|---|---|---|
| Discap | Server-native meeting notes: auto-joins, posts an AI recap with decisions and action items in the channel, and lets you ask past calls — all free, with recording announced and per-user opt-out. | You mainly want translation into many languages or live word-by-word captions — different jobs; Discap is built for the recap and asking past calls. |
| Craig | Podcasts and actual-play shows that want editable per-speaker audio files to cut later. | You want text in the server. The free tier posts no recap and no transcript — just a 7-day audio link. |
| DiscMeet | Teams that want an AI recap plus an MCP bridge into ChatGPT/Claude and don't mind paying for it. | You want recording announced and free auto-join. Consent is opt-in and off by default, so it can transcribe a room silently; auto-join and MCP are Pro. |
| NoteCat | Communities that want a live in-channel transcript plus a web dashboard, free. | You need free /ask or free auto-join (both are $7.99 Premium), or you can't tolerate a recording that occasionally hangs in "processing". |
| NotesBot | Buyers who like usage-based pricing and a Google Docs export of every call. | You want an ongoing free tier — there's only a one-time 30-minute trial before you pay. |
| Scripty | Free live accessibility captions across many languages, "set it and forget it". | You want a post-call recap, summary, or stored transcript. The session ends with "Left VC successfully." and no artifact on free. |
| Scriber | A dead-simple, free, genuinely live transcript posted into the channel as people talk. | You want memory. Nothing is stored, so there is no ask-your-calls, no history, and no recap. |
| iTranslator | Large multilingual communities that mainly need text translation at scale. | You want voice on any budget — every voice feature is locked behind the €24.99/mo Pro tier. |
Quick picks by use case
If you would rather not read the whole matrix, start by naming the job. That decides the category before you test any single tool.
| Use case | Category to test first | Why |
|---|---|---|
| Meeting notes from a Discord voice call | An AI meeting-note bot like Discap | You need a recap with decisions and action items posted in the server, plus the ability to ask the call later — not just raw text. |
| Podcast or actual-play audio | A recorder like Craig | Editable multi-track audio matters more than AI notes. See recorder vs transcription bot. |
| Live accessibility captions | A speech-to-text bot like Scripty or Scriber | Caption latency matters more than post-call recap format. See live transcript vs summary. |
| Translation-heavy text community | A translation bot like iTranslator | Translation workflows are a different job from meeting transcription. |
| Transcribing Discord voice messages | A voice-message utility bot | A context-menu voice-message transcriber may be all you need — that is not call transcription. |
Where Discap fits (and where it doesn't)
Discap should not pretend to be every Discord speech tool. Its lane is meeting notes for Discord voice: it auto-joins when a call starts, announces that it is recording, and when the call ends it posts an AI recap — a short summary, decisions, and action items — into your server, then lets anyone ask questions about past calls. Recording is always announced, members can always opt out, recaps don't expire, and the audio is downloadable. All of that is free.
Different tools specialize, so name your job first: a translation-first bot for many languages, a caption bot like Scripty or Scriber for live word-by-word captions during the call, Craig for editable multi-track studio audio, and Discap for an AI recap posted in your server the moment the call ends — decisions, action items, and the ability to ask past calls, all free. The honest version of "best Discord transcription bot" is: name that job, then test one tool against a real call, because a Top.gg rating won't tell you whether the recap is any good.
For deeper one-to-one comparisons, see our Craig alternative, DiscMeet alternative, and NotesBot alternative write-ups, the AI meeting note-takers for Discord overview, and how we handle recording consent and privacy.
Try It In A Real Discord Call
Add Discap to a test server, start a short voice call, and compare the posted summary, decisions, action items, and transcript against the meeting you actually held. One real call tells you more than any listing page.
Sources Checked
Bots were invited and, where possible, voice-tested hands-on in July 2026. Server counts, features, and prices change; verify them in-product before making a buying decision. Competitor accuracy figures are "in our hands-on test," not absolute ratings.