TTS batch generation test on the statbox
ActivePublic
Actions

Authored by kevinbazira on Mon, Sep 14, 12:09 PM.

Project Tags

Referenced Files

	F102909317: TTS batch generation test on the statbox
	Mon, Sep 14, 12:09 PM

Subscribers

None

	$ python3 -m venv venv && . venv/bin/activate && pip install requests
	CANARY=/srv/published/wmf-ml-models/tts/experiment-v1/_canary

	$ cat > canary.json <<'JSON'
	[{"title": "Earth", "page_id": 9228, "rev_id": 1362915217},
	{"title": "Killdeer", "page_id": 276101, "rev_id": 1361766784},
	{"title": "Leg before wicket", "page_id": 286694, "rev_id": 1366875154}]
	JSON

	$ export REQUESTS_CA_BUNDLE=/etc/ssl/certs/ca-certificates.crt

	$ python3 batch_generate.py \
	--dataset canary.json --sections lead \
	--base https://tts-section-generator.discovery.wmnet:31443 \
	--artifact-dir "$CANARY" --manifest-dir "$CANARY" \
	--log ~/tts/canary.jsonl --concurrency 2
	Batch: 3 articles from canary.json (pinned revisions); 0 sections already settled in /home/kevinbazira/tts/canary.jsonl; concurrency 2; artifacts ['audio_mp3', 'captions_vtt']; scope lead; writing artifacts to /srv/published/wmf-ml-models/tts/experiment-v1/_canary
	3 sections to generate


	Done: 3 ok, 0 skip, 0 fail in 2.4 min
	Articles: 3 manifests written, 0 incomplete (dead letter: fail records in /home/kevinbazira/tts/canary.jsonl), 0 with no generatable sections


	$ find "$CANARY" -type f \| sort && ls -la "$CANARY/enwiki/9228/1362915217/"
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/276101/1361766784/lead.mp3
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/276101/1361766784/lead.vtt
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/276101/1361766784/manifest.json
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/286694/1366875154/lead.mp3
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/286694/1366875154/lead.vtt
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/286694/1366875154/manifest.json
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/9228/1362915217/lead.mp3
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/9228/1362915217/lead.vtt
	/srv/published/wmf-ml-models/tts/experiment-v1/_canary/enwiki/9228/1362915217/manifest.json
	total 1516
	drwxr-xr-x 2 kevinbazira wikidev 4096 Sep 14 11:20 .
	drwxrwxr-x 3 kevinbazira wikidev 4096 Sep 14 11:18 ..
	-rw-r--r-- 1 kevinbazira wikidev 1513584 Sep 14 11:18 lead.mp3
	-rw-r--r-- 1 kevinbazira wikidev 24193 Sep 14 11:18 lead.vtt
	-rw-r--r-- 1 kevinbazira wikidev 626 Sep 14 11:20 manifest.json

	$ python3 -m json.tool "$CANARY/enwiki/9228/1362915217/manifest.json"
	{
	"schema_version": 2,
	"scope": "lead",
	"wiki_id": "enwiki",
	"page_id": 9228,
	"rev_id": 1362915217,
	"generation_version": "kokoro-v1.0+af_heart+norm-2026.07.23-nemo-98d86449",
	"generated_at": "2026-09-14T11:20:08+00:00",
	"sections": [
	{
	"section_id": "lead",
	"title": "Lead",
	"level": 1,
	"duration_ms": 252201.7,
	"content_sha256": "457971c72fa3de06f9d9a44f4e1db345e59f5c6e5a92df0459376560f94f15fe",
	"audio": {
	"key": "enwiki/9228/1362915217/lead.mp3",
	"media_type": "audio/mpeg"
	},
	"captions": {
	"key": "enwiki/9228/1362915217/lead.vtt",
	"media_type": "text/vtt"
	}
	}
	]
	}

Event Timeline

kevinbazira created this paste.Mon, Sep 14, 12:09 PM

kevinbazira mentioned this in rMLISc31de1ee7d3d: tts-section-generator: prepare batch generation script for initial run.Mon, Sep 14, 1:37 PM

TTS batch generation test on the statboxActivePublicActions

Event Timeline

TTS batch generation test on the statbox
ActivePublic
Actions