LIVE
Architectural tweaks may break conventional scaling law exponents16/09/26|OpenAI publishes a framework for reporting model misalignment16/09/26 · OpenAI|NVIDIA's Vera Rubin NVL72 Debuts in MLPerf Inference v6.116/09/26 · NVIDIA|OpenAI expands ChatGPT advertising with Sponsored Agents16/09/26 · OpenAI|OpenAI moves into advertising with 'Sponsored Agents'16/09/26 · OpenAI|Google DeepMind Introduces Gemini 3.8 Live and Its Extended Thinking Variant15/09/26 · Google DeepMind|What's at stake in AI's trillion-dollar infrastructure bet15/09/26|A Flaw in Chain-of-Thought Safety Monitoring14/09/26|Stellar Colosseum: A Multi-Agent System for Long-Horizon Mathematical Research14/09/26|Apple Code Hints Siri Could Be Swapped for ChatGPT or Claude14/09/26 · Apple|Anthropic says Houthi-linked actors used Claude Code for missile guidance software13/09/26 · Anthropic|Yoshua Bengio examines why AI agents lie, cheat and coordinate13/09/26|Architectural tweaks may break conventional scaling law exponents16/09/26|OpenAI publishes a framework for reporting model misalignment16/09/26 · OpenAI|NVIDIA's Vera Rubin NVL72 Debuts in MLPerf Inference v6.116/09/26 · NVIDIA|OpenAI expands ChatGPT advertising with Sponsored Agents16/09/26 · OpenAI|OpenAI moves into advertising with 'Sponsored Agents'16/09/26 · OpenAI|Google DeepMind Introduces Gemini 3.8 Live and Its Extended Thinking Variant15/09/26 · Google DeepMind|What's at stake in AI's trillion-dollar infrastructure bet15/09/26|A Flaw in Chain-of-Thought Safety Monitoring14/09/26|Stellar Colosseum: A Multi-Agent System for Long-Horizon Mathematical Research14/09/26|Apple Code Hints Siri Could Be Swapped for ChatGPT or Claude14/09/26 · Apple|Anthropic says Houthi-linked actors used Claude Code for missile guidance software13/09/26 · Anthropic|Yoshua Bengio examines why AI agents lie, cheat and coordinate13/09/26|
Practical recipe · Audio, voice & music

Transcribe and subtitle a video without uploading it anywhere

A clean SRT subtitle file, generated on your machine, ready for YouTube, LinkedIn or your editor.

Beginner 30 min

The tools in this recipe

Step by step

  1. 1

    Pick the model

    In Buzz, choose the “medium” model for French: “small” gets proper nouns wrong, “large” is three times slower for a small gain on a clear voice.

  2. 2

    Run the transcription

    Drop the video, set the language by hand (auto-detect fails on English intros). A 20-minute video takes 5 to 15 minutes depending on your machine.

  3. 3

    Fix in the editor

    Reread while listening at 1.5×. Fix proper nouns, acronyms and numbers: those are the three usual error sources.

  4. 4

    Export as SRT

    Export as SRT with segments of 2 lines maximum. A 3-line subtitle is unreadable on a phone.

  5. 5

    Translate if needed

    For an English version, run the SRT through DeepL keeping the numbers and timecodes: they are not translated.

Real example

Situation
22-minute internal webinar, one voice, French, 480 MB MP4 file.
What we asked / set
No prompt: Buzz is a local tool without a chat. Settings used: Whisper medium model, French forced, SRT export.
What we got
Transcription in 9 minutes on a 16 GB laptop without GPU. 31 manual fixes over 2,900 words, almost all on product names. Final SRT of 214 segments.

Check by hand before using it

  • Proper nouns and acronyms are spelled as on your materials.
  • No subtitle exceeds 2 lines or 42 characters per line.
  • The first and last segments are in sync (drift shows there first).

Your data: precautions

  • !The file never leaves your machine with Buzz or MacWhisper: that is the point of this recipe. A “free” online service often keeps a copy.
  • !If the video shows people, the transcript contains their words: same retention rules as the video.
  • !The Whisper “large” model needs 10 GB of memory; on a tight machine it crashes without a clear message.

Share this verdict

One link, the same shortlist, the same reasons.

Each recipe was actually done. The tools cited are scored and rechecked every week; no placement is sold.

All recipes