{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-08-18-004",
    "source_story_id": "tmp-story-latent-to-pixel-diffusion",
    "edition_id": "mp-2026-08-18-morning-0040",
    "edition_url": "https://themachinepress.com/edition/2026-08-18",
    "position": 4,
    "story_type": "dispatch",
    "section": "media-creative-tools",
    "editorial_classification": "editorial",
    "headline": "The Image Model Learned in Latents, Then Finished in Pixels",
    "slug": "the-image-model-learned-in-latents-then-finished-in-pixels",
    "dek": "A latent-to-pixel training recipe matched or beat latent diffusion baselines while cutting reported inference time by 3.18 to 4.75 times.",
    "summary": "A latent-to-pixel training recipe matched or beat latent diffusion baselines while cutting reported inference time by 3.18 to 4.75 times.",
    "body_text": "Direct large-scale pretraining in pixel space converged much more slowly than latent-space training in the authors’ study. Their recipe first acquires generative priors in latent space, then moves to pixels during post-training while tuning initialization, data mix, prediction target, decoder and noise schedule. The resulting models matched or outperformed the tested latent counterparts and delivered reported end-to-end speedups of 3.18 to 4.75 times. Those gains belong to the evaluated architectures and training setup, not every text-to-image system.",
    "why_it_matters": "A latent-to-pixel training recipe matched or beat latent diffusion baselines while cutting reported inference time by 3.18 to 4.75 times.",
    "limitations": [
      "Those gains belong to the evaluated architectures and training setup, not every text-to-image system."
    ],
    "importance": 9,
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-18-004/the-image-model-learned-in-latents-then-finished-in-pixels",
    "json_url": "https://themachinepress.com/story/mp-2026-08-18-004.json",
    "first_published_at": "2026-08-18T09:00:00.000-04:00",
    "modified_at": "2026-08-18T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-08-18-004-001",
        "text": "A latent-to-pixel training recipe matched or beat latent diffusion baselines while cutting reported inference time by 3.18 to 4.75 times.",
        "source_ids": [
          "source-2026-08-18-004"
        ],
        "qualification": "Those gains belong to the evaluated architectures and training setup, not every text-to-image system."
      }
    ],
    "source_ids": [
      "source-2026-08-18-004"
    ],
    "tags": [
      "diffusion",
      "pixel space",
      "image generation",
      "inference"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-08-18-004",
      "title": "arXiv preprint 2608.16887",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2608.16887",
      "canonical_url": "https://arxiv.org/abs/2608.16887",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-08-16T20:00:00.000-04:00",
      "accessed_at": "2026-08-18T08:25:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-08-18-004-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The Image Model Learned in Latents, Then Finished in Pixels",
    "publisher": "The Machine Press",
    "published_at": "2026-08-18T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-08-18-004/the-image-model-learned-in-latents-then-finished-in-pixels"
  }
}