What audio format does Discord use for voice chat?

Discord uses the Opus codec at 48 kHz for every voice call, video call, screen share with sound, and stage channel — there is no setting to change this. The bitrate scales with the server's boost level: 64 kbps default, 96 kbps once a server admin raises it, 128 kbps at Boost Level 1, 256 kbps at Level 2, and 384 kbps at Level 3. WebRTC adapts downward automatically if your network is poor. Local file format is irrelevant once you are in a voice channel.

What is the file size limit for audio on Discord?

Free accounts can upload up to 25 MB per file (raised from 8 MB in April 2024). Nitro Basic is 50 MB, full Nitro is 500 MB. Server boosts also raise the per-server ceiling: Boost Level 2 lifts everyone on that server to 50 MB and Level 3 lifts to 100 MB. These limits apply per attachment, so splitting a long recording into multiple smaller MP3s is a viable workaround on the free tier.

Can I share FLAC files on Discord?

Yes — FLAC plays in Discord's inline player on web, desktop, and both mobile clients. The catch is size: FLAC averages around 5–6 MB per minute for 16-bit/44.1 kHz stereo, so a 4-minute song is already 25 MB. Free-tier users will run out of upload room quickly. If preserving lossless quality matters, FLAC is fine; if you just want the recipient to hear the music, MP3 at 192 kbps holds roughly four times as much audio per megabyte.

Why does music sound worse when I screen-share with sound?

Screen-share audio routes through the same Opus voice pipeline that microphone audio uses, so it is re-encoded to the channel's bitrate (typically 64–128 kbps) before reaching listeners. A 320 kbps source MP3 ends up sounding like 64 kbps Opus on the other end. For music listening parties, upload the file as an attachment and have everyone press play locally — Discord's player decodes the original file directly with no re-encode.

What is the best format for a Discord music bot to play?

Bots that use discord.js or discord.py expect either a raw Opus stream or PCM that the library will Opus-encode internally via FFmpeg. The cleanest input is 48 kHz stereo, since Discord runs at 48 kHz natively and any other rate triggers a resample. Pre-converting source files to 48 kHz Opus removes that step entirely and reduces CPU load on lightweight hosts. MP3, M4A, WAV, and FLAC all work, but they all funnel through FFmpeg before reaching the voice gateway.

Best Audio Format for Discord: Opus, MP3, and File Limits

Discord splits audio into two distinct pipelines: live voice (always Opus) and uploaded audio attachments (whatever you send, played in the inline player). Picking the right format means understanding which pipeline you are touching, what the upload limit is on your tier, and which container formats Discord's player will actually decode on each client.

Discord Voice Channels Use Opus Internally

Every voice call, video call, and screen-share session on Discord encodes microphone audio as Opus before it leaves your machine. The default voice channel bitrate is 64 kbps. Server admins can raise it to 96 kbps on a standard server, 128 kbps once the server hits Boost Level 1, 256 kbps at Level 2, and 384 kbps at Level 3. Stage channels follow the same scaling. You do not pick a codec or a sample rate from the client — Discord handles encoding at 48 kHz and lets WebRTC adapt the bitrate down toward 8 kbps if your network gets congested.

That means uploaded file format is irrelevant for live conversations. The variables that affect call quality are microphone hardware, the server's boost tier, the Krisp noise suppression toggle in Voice and Video settings, and whether your ISP throttles UDP. If a friend complains your voice sounds compressed, raise the server boost tier or check that the Audio Subsystem is set to Standard rather than Legacy.

Upload Size Limits by Tier

Discord raised the free upload limit in 2024. Current ceilings:

Free accounts: 25 MB per file (was 8 MB before April 2024)
Nitro Basic: 50 MB per file
Nitro: 500 MB per file
Server Boost Level 2: raises everyone's limit on that server to 50 MB
Server Boost Level 3: raises to 100 MB

These limits are per-attachment. The 25 MB free tier holds about 26 minutes of 128 kbps MP3, 17 minutes of 192 kbps MP3, or 10 minutes of 320 kbps MP3. For anything longer, host on Google Drive, Dropbox, or SoundCloud and paste the link — Discord unfurls most public audio URLs into an inline player automatically.

Formats Discord's Inline Player Decodes

Discord's web, desktop (Electron), iOS, and Android clients all use the same player layer for attachment playback. Tested formats and their behavior:

MP3 — embeds and plays everywhere. The safest default.
OGG Vorbis — embeds and plays on every client. Better quality than MP3 at equal bitrate.
OGG Opus / .opus — plays on desktop and mobile; some older Android builds fall back to download.
M4A (AAC in MP4 container) — plays everywhere. iPhone Voice Memos export this directly.
FLAC — plays inline. Files are large; 4 minutes of stereo FLAC is roughly 25 MB.
WAV — plays inline but eats your upload quota. 5 minutes of CD-quality WAV is about 50 MB.
AIFF, WMA, AC-3 — usually shown as plain attachments without an inline player. Recipients have to download.

Recommended Upload Format

For voice memos, podcast clips, and most content, MP3 at 192 kbps stereo is the right balance — universally decodable, small enough that a 25 MB ceiling holds 17 minutes, and indistinguishable from higher bitrates on Discord's playback layer. Drop to 128 kbps mono for spoken-word content; you'll fit roughly 26 minutes per attachment and the speech intelligibility is identical.

For music sharing where the recipient might want to keep the file, OGG Vorbis at q5 or q6 (roughly 160–192 kbps VBR) sounds noticeably cleaner than MP3 at the same size. For lossless masters going to a producer or mastering engineer, send FLAC if the file fits — it preserves bit-exact audio and Discord's player will preview it.

Screen-Share and Stage Audio Behavior

Screen-share with the "Share Sound" toggle pulls system audio through the same Opus pipeline as voice. If you play a 320 kbps MP3 from your local machine while sharing, listeners hear it re-encoded to whatever bitrate the channel is set to — typically 64–96 kbps. Quality drops accordingly. For high-fidelity music listening parties, upload the file as an attachment and have everyone play it locally instead.

Stage Channels (used for community events) cap at 64 kbps Opus regardless of server boost level. They are designed for moderated talks, not music broadcast.

Music Bots and Custom Bot Audio

Public music bots like Hydra, Jockie Music, and FredBoat stream from YouTube, Spotify, or SoundCloud and never touch your local files. Quality is bounded by the source stream and the channel bitrate.

If you are writing a custom bot with discord.js or discord.py, the voice library expects either a raw Opus stream or PCM that it will Opus-encode for you. Both libraries shell out to FFmpeg by default. Common gotcha: feeding a 44.1 kHz file when Discord wants 48 kHz triggers a resample inside FFmpeg that some hosts (Replit, small VPS) cannot keep up with — pre-convert your audio to 48 kHz stereo before deployment. The WAV to Opus converter produces files that drop straight into discord.js play streams.

Converting Audio to Fit Discord

Three common scenarios:

Trim a large WAV to fit 25 MB free: convert with WAV to MP3 at 128 kbps mono, or run an existing file through the audio compressor to slide under the cap without changing format.
Send a high-quality music clip to a Nitro friend: WAV to FLAC keeps it lossless under the 500 MB limit.
Share an iPhone voice memo: the M4A plays as-is. If you need MP3 for a non-Discord recipient, run M4A to MP3.

Voice Channel Bitrate Adjustments

Server admins can change voice channel bitrate from the channel's edit menu (right-click channel > Edit Channel > Bitrate). The slider adjusts in 4 kbps steps. Higher bitrates use more bandwidth on every participant's connection — 384 kbps stereo on a 5-person call burns roughly 1.9 Mbps of upstream per speaker. For most communities, 64–96 kbps is plenty; bumping to 256 or 384 kbps only matters for music groups where listeners use good headphones in quiet rooms.

The Voice & Video user settings on each individual machine include Krisp noise suppression (effective for keyboard, fan, and street noise; not great for music), echo cancellation, and automatic gain control. For musicians, all three should be off; for typical conversation, all three on.

For the broader question of why audio files balloon in size, see why is my audio file so large and the audio bitrate guide by use case.