Start free

Phetolela MP3 go sengwalwa.Mawala a mosaatsanyi, dipuo tše go fapana le 100.

Akoma faele ya MP3 mo go neng ka neng mo 64 go 320 kbps. Amogetsa phetoleling e e nang le nako le mawala a mosaatsanyi mo go dipuo e 99 — ga le tlhoke go badila mokgatlho, ga o re-encode, ga le hemedi mo listeng.

Drop a file, or pick one

MP3 · WAV · M4A · MP4 · MOV · MKV · OGG · OPUS · FLAC · WEBM — up to 100 MB anonymously

Paste a link, we’ll fetch the audio

YouTube · TikTok · Vimeo · Twitter · SoundCloud · Spotify · 50+ more

Record straight from your browser

Sign up takes 30 seconds — recording opens right after, in the dashboard.

No card required~90s per 60-min fileSRT · VTT · DOCX · TXTFiles auto-deleted in 24h

↓ Leba seleo se se etsang

MP3 e fela. Phetoleling e e arogantsweng.

Re balela diretswa tša diframe tša MP3 ka go ithe — VBR, CBR, joint-stereo, encoder neng (LAME, Fraunhofer, FFmpeg). Fa faele e le stereo e e boiteanong go lebakabaka le ba mosaatsanyi go le different channels, re dirisa seo go go rarola diphathi. Mono mix-down e wa go fallback go diarization e e nang le kumo.

interview-tape-04.mp3REC 192 kbps · stereo · 38:42
auto-detected en-GB44.1 kHz · LAME 3.100
~90s
Phetoleling · go gobala95% go nepo
S1

So when did you first realise the archive was incomplete?

S2

Probably around 2019, when we started digitising the reel-to-reels.

S1

And the missing tapes — were they catalogued anywhere at all?

S2

There's a paper index from '78, but half of it's water-damaged.

95% mo 192 kbps stereoSRT · DOCX · TXT · JSON · VTT

↓ This is the dashboard

This is what loads when the job finishes.

Same layout as the real dashboard — Summary, full Transcript, Speakers tab, Exports. Key points and action items extracted automatically. Auto-tags on every job.

Try it on your own file — it's free

Dikgetsi tše tharo tše di mo gare · kutlo ya boitshoko

Whisper e e boikgang mo locally. Otter kgotsa Sonix. Kgotsa ona.

O ka diranela Whisper mo lapeng la gago fela fa o bo technical. Otter le Sonix di amogetsa MP3 uploads mo go dashboards tša subscription. Ree amogetsa faele, re go bega phetoleling, le ga o ka gona mo UI.

Option 01

Whisper local / open source

Boikgang fa o na le GPU le letsatsi le lengwe. Ga go na diarization ya mosaatsanyi mo go go akoya.

Go simololaPython + CUDA + 10 GB models
Diarization ya mosaatsanyiGa se seano (pyannote add-on)
Lebelo · hora e 1 ya MP35–40 min mo GPU e e se ye mosebeletsha
Dipuo99, mme modele o motelele o wa se ka fapane go tlase ga 80%
ExportTXT / SRT / VTT / JSON
DitshwaneloBoikgang + motlakase wa gago
Best forDipallase tšeo di na le GPU, ga di tlhoke mawala a mosaatsanyi, le di rata go durišetsa kwa ntle.
Option 02

Transcription.Solutions

Akoma MP3. Amogetsa sengwalwa se se nang le mawala a mosaatsanyi ka nako e e swanetseng.

Go simololaDrag-and-drop, ga le tlhoke akhaunto go go leka
Diarization ya mosaatsanyiE a agilwe (Pro & Business plans)
Lebelo · hora e 1 ya MP3~90 sekunde
Dipuo99, auto-detected
ExportSRT · VTT · DOCX · TXT · JSON
Ditshwanelo · per min$0.03
Best forMotho yo mongwe le yo mongwe yo o nang le MP3 — lekolete la mojournalist, podcast export, voice memo, archival dub — yo o mo go o eletsa sengwalwa se se nepelo fela.
Option 03

Otter / Sonix

Dashboard e e lokileng, monthly minutes cap, English-tuned. File upload e utwa jaaka setheo se se sekondari.

Go simololaAkhaunto + phalano e e diragatsweng
Diarization ya mosaatsanyiAcoustic, EN-leaning
Lebelo · hora e 1 ya MP35–10 min mo leila
DipuoOtter EN-only; Sonix ~40
ExportGo kwelobagile mo diphalano tšee di diragatsweng
Ditshwanelo$17+/mo kgotsa $10+/hr (Sonix)
Best forDithuto tšee di batšwakgetse transcript editor le collaboration UI go feta faele→sengwalwa flow e e nepelo.

Pricing le feature availability di nepele go ya go May 2026. Whisper performance e fapanika go ya go model size le hardware.

E le seethemogo sa MP3

Dinto tše tharo tšee di golaya batho mo go tools tšee di se fetogo.

MP3 ke mokgatlho, e se recording style — go duma seleo sa go botlhokwa go tswa mo encoder, e se puo.

Seelo se se dirago mathata

  1. 1VBR headers di arogana ka go makala. Tools e mengwe e balela variable-bitrate MP3s jaaka fixed-rate le e loga duration go phoso — timestamps di fapala mo metsotso mo hora e kakaretso.
  2. 2Joint-stereo e bolawa go mono go upload preprocessing. O lošwa le per-speaker channel separation e e neng le fela mo faele.
  3. 3Embedded ID3 album art e bolaya a le mang uploaders — ba lala faele jaaka 'not pure audio' kgotsa ba e šoša le go re-encode, go lošwa quality e le morago.

Se re se dirang gone

  1. 1Re dirišetša Xing/LAME header fa se dutegang le frame-count fallback fa se se na. VBR timestamps di tšeala accuracy go ±0.1 s mo dihora tšee di le ntši.
  2. 2Joint-stereo le true-stereo MP3s di decoded go L/R PCM before diarization. Fa ba mosaatsanyi ba ne ba paned, re di tšiaba di le aparate.
  3. 3ID3v1, ID3v2, APE tags, embedded art — esela e gatetswe ka di-untouched. Ree re-encode go MP3 ya gago.

Recommended job settings go MP3 uploads

Defaults tšee di swanela ~80% ya mafaele a MP3. Override per-job go forma.

Decoder
Frame-accurate, no re-encode
Diarization
Channel split fa stereo, else acoustic
Speaker model
Auto · 1-12 speakers
Puo
Auto-detect go tswa go selekanyo se go 30 s
Filler words
Removed (toggle go go bolaya)
Export bundle
DOCX + SRT + timestamped TXT

Accuracy · real-world numbers

95%+ mo 192 kbps stereo. E ka dirišega go 64 kbps mono.

MP3 accuracy e kopantšwa ke seelo se compressed keeper e sa se gatile, e se ke a. Perceptual compression mo go feta ~96 kbps e sireletša intelligibility ya puo gabotlane; tlase ga 64 kbps, sibilants le consonants ba simolola go bolawa. Dinomoro tlase e tšwa mo MP3s e e bonolo ya customer mo go go dirago tikologo.

96%
320 kbps stereo, studio source

Go kakaretso-lossless go puo. Podcast masters, dictation app exports, professional interview rigs. Diarization e nepelo fa ba mosaatsanyi mo separate channels.

95%
192 kbps stereo, 2-3 speakers

Bitrate e e lešoro ka kakaretso go MP3s tšee di batlangaang puo. Zoom exports, Riverside downloads, voice recorders e simolola mo default. Compression artifacts ga di utolole go recognizer.

91%
128 kbps mono, conversational

Voice memo defaults mo difoneng tšee di le ntši. Acoustic diarization e lametša 2-4 speakers. Dinomoro le proper nouns ka nako e nngwe di tlhoka leba.

84%
64 kbps mono, archival / phone-dump

Old answering-machine rips, lecture archives, narrow-band sources. High-frequency consonants (f/s/sh) ba bolawa. Le boela ba legible — beakanya proofread.

Dipotšo tšee di e kgonegile

Dinto e 8 tšee ba go botša ka phetolelo ya MP3.

01Ke bitrate e e lowane ya MP3 e e tšwanelang gore le amogetse sengwalwa se se sebelang?+
64 kbps ke floor e e go botlhokwa. Go tlase ga seo, sibilants (s, sh, f) di komparesa go noise le word error rate e gopela go feta 20%. Fa o bo o rekota go mosa, tlo go 128 kbps mono kgotsa 192 kbps stereo — neng e nngwe e le go metlala go puo.
02A ke tlhoka go badila MP3 ya me go WAV first?+
Aowa. Re-encoding MP3 → WAV ga o akoma lefeela tšeo dicuracy tšee kuta data e encoder e se gatile e boela e bolelele botsona. Apaya MP3 ka go ithe. Re decode frames mo memoring le re feed PCM go recognizer.
03A stereo MP3 e go ba le better speaker labels go feta mono?+
Fela fa ba mosaatsanyi ba ne ba rekotwa mo separate channels — most stereo MP3s di na le audio e e swanang mo diphapogo peli ('dual mono') le ga ba na le bothata. True channel-split (e.g. Riverside exports, two-mic field rigs) e re dumela go omela acoustic diarization le go label ba mosaatsanyi go ka kakaretso-perfectly.
04Ke saizi e e kgethegileng ya MP3 file e o amogelanang?+
5 GB per upload, e e gorang ~60 hours mo 192 kbps kgotsa 90 hours mo 128 kbps. Fa faele ya gago e le kgolo e go feta re go beka chunked upload — ga o tlhoke go go rarola gago.
05Hora e 60-minute ya MP3 e amogetsa gore le transcrebe metsotso e mesomme?+
Metsotso e 90 go tswa go upload-complete go transcript-ready, e le neng bitrate. Decoding MP3 frames e potlakile; nako e mo recognizer. Diarization e oketša metsotso e 5-10 mo mafaele a multi-speaker.
06MP3 ya me e na le mosaano wa background music — a phetoleling e tla bolawa?+
Quiet bed music tlase ga puo e nepelo. Loud music e e lokanang le voice (intro stings, scoring under interviews) ka nako e nngwe e ka bolela misrecognitions mo overlapping syllables. Toggle music suppression mo job form go pre-filter.
07A o ka dirišetša MP3s tšee di ripped go tswa mo phone voicemail kgotsa answering machines?+
Eya, le fa dinnye di le dingwe tšea ke 8 kHz narrow-band re-encoded jaaka MP3 — audio quality ceiling e thakisa ke origenal PSTN capture, e se MP3 wrapper. Ela accuracy ya 78-85% mo go genre e, e le seo sa underlying call.
08A o bolaya MP3 ya me morago ga phetoleling e e fellwa?+
Mafaele a bolawa after 30 days mo default, kgotsa ka go ithe on request via dashboard. Phetoleling e tšea mo account ya gago go go fiwa o e tla bue. Ree ga re dirišetša customer audio go go akela modele neng — nnga.

Akoma MP3 ya gago. Amogetsa sengwalwa go 90 sekunde.

Metsotso e 30 e le boikgang mo kgwedi e nngwe. Ga le tlhoke karata. Mawala a mosaatsanyi, dipuo tše 99, export format nngwe le nngwe e akantswe.

Simolola boikgang