Open-Source Descript Alternative (2026, Transcription)
Identifying Descript · Mark source
This hub lists open speech-to-text people compare with Descript-class transcript editors. Descript is named only to identify that product class. The sourced engine today is whisper.cpp. We do not rank transcription or overdub quality against Descript.
- whisper.cpp — MIT C/C++ port of OpenAI Whisper for local speech-to-text.
- Granite Speech 5.0 — Apache-2.0 IBM speech family — 5.0-2B ASR/AST plus 4.1-2B Plus speaker attribution.
- Nemotron 3.5 Transcribe — NVIDIA offline speech MoE (30B-A4B; 34B / 4.3B active). OpenMDW-1.1. Not Bolt or Super VL.
How we pick
- Official repository we could open.
- License copied from that host.
- Local transcription or speaker turns — not an unsourced NLE list.
- No unsourced “beats Descript Overdub” claims.
Comparison
| Name | License | Stars (GitHub) | Platforms | Runs locally | Key strengths |
|---|---|---|---|---|---|
| whisper.cpp | MIT | 54,489 | macOS, Windows, Linux, iOS, Android | Yes | MIT C/C++ port of OpenAI Whisper for local speech-to-text. |
| Granite Speech 5.0 | Apache-2.0 | See page | See model card | Yes | Apache-2.0 IBM speech family — 5.0-2B ASR/AST plus 4.1-2B Plus speaker attribution. |
| Nemotron 3.5 Transcribe | other | See page | See model card | Yes | NVIDIA offline speech MoE (30B-A4B; 34B / 4.3B active). OpenMDW-1.1. Not Bolt or Super VL. |
Lead engine: whisper.cpp. File ASR/AST: Granite Speech 5.0 (IBM Apache-2.0 ~2.3B) and Nemotron 3.5 Transcribe (NVIDIA OpenMDW-1.1 30B-A4B; not Bolt/Super VL) — Nemotron 3.5 Transcribe vs Granite Speech 5.0. Simultaneous speech MT (not a timeline editor): Hy-SimulMT — Hy-SimulMT vs Granite Speech 5.0 — and Hibiki-Zero (fr/es/pt/de→en) — Hibiki-Zero vs Hy-SimulMT. Speaker turns: pyannote community-1 on Otter.ai alternatives. TTS (the other direction): ElevenLabs alternatives (Drama 3, GPT-SoVITS). Avatars: HeyGen alternatives. Video generation: Runway alternatives.
Sourced snapshot (2026-10-19)
| Field | whisper.cpp |
|---|---|
| Repository | ggml-org/whisper.cpp |
| Stars / forks | 54,489 / 6,270 |
| License | MIT |
| Last push | 2026-10-19 |
| Job | Local speech-to-text |
| Quality vs Descript | Not independently tested |
How to choose
- Need local transcription: whisper.cpp.
- Need Apache-2.0 multilingual ASR/AST (or speaker tags on the Plus sibling): Granite Speech 5.0.
- Need NVIDIA offline 30B-class ASR (OpenMDW-1.1): Nemotron 3.5 Transcribe.
- Need live Chinese–English simultaneous speech (not a Descript timeline): Hy-SimulMT.
- Need live French / Spanish / Portuguese / German → English S2ST: Hibiki-Zero.
- Need who spoke when via a diarization pipeline: pyannote community-1.
- Need voice generation: ElevenLabs alternatives.
- Need a hosted studio: that is Descript; this hub does not clone it.
Need a screen recorder, not a transcript editor: Loom alternatives (Cap, OBS Studio).
See also Video & Audio.
Listed options
whisper.cpp
MIT C/C++ port of OpenAI Whisper for local speech-to-text.
Granite Speech 5.0
Apache-2.0 IBM speech family — 5.0-2B ASR/AST plus 4.1-2B Plus speaker attribution.
Nemotron 3.5 Transcribe
NVIDIA offline speech MoE (30B-A4B; 34B / 4.3B active). OpenMDW-1.1. Not Bolt or Super VL.
FAQ
Is there a free open-source Descript alternative?
whisper.cpp (ggml-org/whisper.cpp) is a MIT C/C++ Whisper port with 54,489 stars as of 2026-10-19 and a same-day push. It transcribes locally. It does not replace Descript’s timeline, overdub, or studio UI. We have no WER bake-off.
What about speaker labels?
See pyannote community-1 on Otter.ai alternatives. That is diarization, not an editor.
Need TTS instead of STT?
ElevenLabs alternatives (Drama 3, GPT-SoVITS). Opposite direction of Descript-style transcription.
Need a timeline editor?
Premiere/CapCut alternatives names Kdenlive and Shotcut in prose; those tool pages are not in this batch. Do not treat this hub as an NLE list.
Sources & verification
- https://www.descript.com/
- https://github.com/ggml-org/whisper.cpp
- https://huggingface.co/pyannote/speaker-diarization-community-1
- https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B
License notes are not legal advice. Read the original LICENSE. Trademark & disclaimer.