File formats

Supported input

Audio

.wav, .mp3, .flac, .m4a, .ogg, .opus, .aac, .wma

Video

.mp4, .mkv, .avi, .mov, .webm, .m4v, .flv, .wmv

Audio is extracted with ffmpeg at 16 kHz mono PCM (required by NeMo).

---

Output per job

File	Format	Content
trascrizione.txt	Plain UTF-8 text	Full transcript
sottotitoli.srt	SubRip	Segments with timestamps
riassunto.txt	Plain UTF-8 text	Summary (if generated)
source.*	Original	Copy of the uploaded file
job.json	JSON	Job metadata

---

SRT format

Standard SubRip:

1
00:00:00,000 --> 00:00:05,120
First transcribed sentence.

2
00:00:05,120 --> 00:00:10,450
Second sentence.

Generated by src/sbobinator/export.py from NeMo timestamps (when available).

---

Long files

Threshold	Behavior
≤ 30 minutes	Full transcription
> 30 minutes	30 s chunks with 2 s overlap, then merge

Configurable in TranscribeConfig (chunk_threshold_sec, etc.).

---

Legacy output (--legacy-output)

Without a job folder:

data/output/
├── nomestem.txt
├── nomestem.srt
└── nomestem_riassunto.txt

May overwrite files with the same name. Prefer the default job mode.

---

Encoding

All text files: UTF-8.