{
  "formatVersion": 1,
  "scope": "Curated public research extract; not the full raw run.",
  "canonical": "https://embiro.ai/research/shared-media-on-one-t4",
  "fieldNote": 9,
  "slug": "shared-media-on-one-t4",
  "title": "How much media fits on one T4?",
  "description": "Speech, images and a small video experiment reveal where a compact shared worker starts to strain.",
  "category": "Infrastructure",
  "art": "vision-lens",
  "thumbnail": {
    "src": "/research/thumbnails/shared-media-on-one-t4.webp",
    "alt": "A lens, waveform and film frame balanced above one compute pedestal."
  },
  "date": "2026-10-11",
  "published": "2026-10-12",
  "updated": "2026-10-12",
  "author": "Embiro Labs Research",
  "status": "Exploratory benchmark",
  "headline": "Memory is a product constraint.",
  "metric": "9",
  "metricLabel": "failed shared SDXL requests retained",
  "method": [
    "Cape Town g4dn.xlarge: one T4, four vCPUs, 16 GiB RAM. Gemma 4 E4B, Qwen retrieval and Whisper shared the host. SD1.5 and SDXL were tested one at a time alongside that core.",
    "Images used four matched prompts, seed 42, 25 steps and FP16. SD1.5 ran at 512px; SDXL used offload/slicing/tiling at 512px and 1024px. Core/media mixes used four concurrent clients on loopback.",
    "Whisper evaluated twelve Swahili FLEURS and twelve Nigerian English recordings, with an isolated supplement of six Kenyan, six South African and three Ghanaian clips from an EdAcc derivative. Kokoro used twelve synthetic English generations. Wan video ran separately with core workers stopped."
  ],
  "sections": [
    {
      "title": "What the compact host handled",
      "paragraphs": [
        "SD1.5 completed its four image-quality prompts at a 5.14-second median; two mixed-load images took a 7.95-second median. Adding image work raised queueing for other tasks. These are small workloads, not steady-state capacity estimates.",
        "The shared SDXL attempt exhausted host RAM and its worker was killed. Nine failed requests remain in the evidence. A later isolated SDXL retest cannot be used to claim safe shared capacity."
      ]
    },
    {
      "title": "Audio needs its own evaluation",
      "paragraphs": [
        "Whisper WER was 38.0% on the twelve Swahili clips and 19.1% on twelve Nigerian English clips. Regional acoustic supplements were tiny convenience samples, with differing recording conditions; they do not support ranking accents. Kokoro generated English on CPU with a 2.33-second median across twelve requests."
      ]
    },
    {
      "title": "Feasible is not production-ready",
      "paragraphs": [
        "Wan 2.1 1.3B produced a 17-frame, 448×256 video at eight frames per second in 258.11 seconds, after 260.98 seconds of readiness work. It required an isolated run and heavy swap. This was a feasibility result, not a usable interactive video service.",
        "Four corrected durable media jobs survived an API restart and idempotent replays. Earlier harness failures are retained; successful queue recovery does not make all co-resident models safe."
      ]
    }
  ],
  "limitations": [
    "Loopback excludes customer networking. Small image sets were inspected qualitatively, with no automated aesthetic score.",
    "Isolated SDXL/video retests do not establish co-residency. GPU sampling may miss brief peaks. Speech samples may overlap model training; no general accent accuracy estimate.",
    "Synthetic billing rates tested arithmetic only. They are not customer pricing."
  ],
  "tables": [
    {
      "title": "Selected observations",
      "columns": [
        "Workload",
        "Result",
        "Boundary"
      ],
      "rows": [
        [
          "SD1.5 quality · 4 images",
          "5.14s median",
          "512px, warm loopback"
        ],
        [
          "SD1.5 mixed · 2 images",
          "7.95s median",
          "Shared host"
        ],
        [
          "SDXL shared",
          "9 failures",
          "Host RAM exhaustion"
        ],
        [
          "Kokoro EN · 12 generations",
          "2.33s median",
          "CPU on shared host"
        ],
        [
          "Wan 1.3B · 1 video",
          "258.11s generation",
          "Isolated; not production capacity"
        ]
      ],
      "note": "Failures and isolated retests are intentionally not merged into one success average."
    }
  ],
  "references": [
    {
      "label": "Google FLEURS · dataset and attribution",
      "url": "https://huggingface.co/datasets/google/fleurs"
    },
    {
      "label": "Kokoro · model card",
      "url": "https://huggingface.co/hexgrad/Kokoro-82M"
    },
    {
      "label": "Nigerian English dataset · Benjamin Ogbonna",
      "url": "https://huggingface.co/datasets/benjaminogbonna/nigerian_accented_english_dataset"
    },
    {
      "label": "EdAcc · Edinburgh CSTR",
      "url": "https://huggingface.co/datasets/edinburghcstr/edacc"
    }
  ],
  "audio": [],
  "evidence": [
    {
      "file": "README.md",
      "sha256": "bdc82eae08967518ad702b964d2d6e3cdad41871317a42fffbc735f4b5f80bbb"
    },
    {
      "file": "summary.json",
      "sha256": "313ed32995503ae058061262e97da7261b2581786a867b2bcb72479c48e9a994"
    }
  ]
}