davanstrien HF Staff Claude Opus 4.7 (1M context) commited on
Commit
d42ff0c
·
1 Parent(s): 51bba34

docs: show URL-based invocation in Quick Start

Browse files

Replace bare filenames with full raw.huggingface.co URLs so
external users can copy-paste without cloning. Add a local-dev
note for contributors working from a checkout.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Files changed (1) hide show
  1. README.md +8 -2
README.md CHANGED
@@ -14,22 +14,28 @@ Scripts for transcribing audio files using HF Buckets and Jobs.
14
 
15
  ## Quick Start
16
 
 
 
17
  ```bash
18
  # 1. Download audio from Internet Archive straight into a bucket
19
  hf jobs uv run \
20
  -v bucket/user/audio-files:/output \
21
- download-ia.py SUSPENSE /output
 
22
 
23
  # 2. Transcribe — audio bucket in, transcript bucket out
24
  hf jobs uv run --flavor l4x1 -s HF_TOKEN \
25
  -e UV_TORCH_BACKEND=cu124 \
26
  -v bucket/user/audio-files:/input:ro \
27
  -v bucket/user/transcripts:/output \
28
- cohere-transcribe.py /input /output --language en --compile
 
29
  ```
30
 
31
  No download/upload step. Buckets are mounted directly as volumes via [hf-mount](https://github.com/huggingface/hf-mount).
32
 
 
 
33
  ## Scripts
34
 
35
  ### Transcription
 
14
 
15
  ## Quick Start
16
 
17
+ Scripts run directly from their Hub URL — no clone or local checkout needed:
18
+
19
  ```bash
20
  # 1. Download audio from Internet Archive straight into a bucket
21
  hf jobs uv run \
22
  -v bucket/user/audio-files:/output \
23
+ https://huggingface.co/datasets/uv-scripts/transcription/raw/main/download-ia.py \
24
+ SUSPENSE /output
25
 
26
  # 2. Transcribe — audio bucket in, transcript bucket out
27
  hf jobs uv run --flavor l4x1 -s HF_TOKEN \
28
  -e UV_TORCH_BACKEND=cu124 \
29
  -v bucket/user/audio-files:/input:ro \
30
  -v bucket/user/transcripts:/output \
31
+ https://huggingface.co/datasets/uv-scripts/transcription/raw/main/cohere-transcribe.py \
32
+ /input /output --language en --compile
33
  ```
34
 
35
  No download/upload step. Buckets are mounted directly as volumes via [hf-mount](https://github.com/huggingface/hf-mount).
36
 
37
+ > **Local dev**: if you've cloned this repo, swap the URL for the local filename (e.g. `cohere-transcribe.py /input /output ...`).
38
+
39
  ## Scripts
40
 
41
  ### Transcription