🐦 Twitter Post Details

Viewing enriched Twitter post

@omarsar0

New research from Google DeepMind. Really interesting paper on diffusion models. Training good latents for diffusion models is harder than it looks. The standard approach uses a KL penalty borrowed from VAEs, with no principled way to control how much information actually lives in the latent space. This new research introduces Unified Latents (UL), a framework that co-trains a diffusion prior on the latents. This provides a tight upper bound on latent bitrate and makes the reconstruction-generation tradeoff explicit and, most importantly, tunable. On ImageNet-512, UL achieves FID 1.4 while requiring fewer training FLOPs than Stable Diffusion latents. On Kinetics-600, it sets a new state-of-the-art FVD of 1.3 for video generation. The latent space is one of the most undertreated design decisions in diffusion-based generation. UL gives practitioners a principled handle on it, for both images and video. Paper: https://t.co/E1HCf9QzB4

Media 1
View on Twitter

📊 Media Metadata

{
  "media": [
    {
      "url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2026701743246254107/media_0.jpg?",
      "media_url": "https://crmoxkoizveukayfjuyo.supabase.co/storage/v1/object/public/media/posts/2026701743246254107/media_0.jpg?",
      "type": "photo",
      "filename": "media_0.jpg"
    }
  ],
  "processed_at": "2026-03-01T19:15:40.985942",
  "pipeline_version": "2.0"
}

🔧 Raw API Response

{
  "type": "tweet",
  "id": "2026701743246254107",
  "url": "https://x.com/omarsar0/status/2026701743246254107",
  "twitterUrl": "https://twitter.com/omarsar0/status/2026701743246254107",
  "text": "New research from Google DeepMind.\n\nReally interesting paper on diffusion models.\n\nTraining good latents for diffusion models is harder than it looks. The standard approach uses a KL penalty borrowed from VAEs, with no principled way to control how much information actually lives in the latent space.\n\nThis new research introduces Unified Latents (UL), a framework that co-trains a diffusion prior on the latents. This provides a tight upper bound on latent bitrate and makes the reconstruction-generation tradeoff explicit and, most importantly, tunable.\n\nOn ImageNet-512, UL achieves FID 1.4 while requiring fewer training FLOPs than Stable Diffusion latents. On Kinetics-600, it sets a new state-of-the-art FVD of 1.3 for video generation.\n\nThe latent space is one of the most undertreated design decisions in diffusion-based generation. UL gives practitioners a principled handle on it, for both images and video.\n\nPaper: https://t.co/E1HCf9QzB4",
  "source": "Twitter for iPhone",
  "retweetCount": 34,
  "replyCount": 8,
  "likeCount": 159,
  "quoteCount": 0,
  "viewCount": 27012,
  "createdAt": "Wed Feb 25 16:52:04 +0000 2026",
  "lang": "en",
  "bookmarkCount": 142,
  "isReply": false,
  "inReplyToId": null,
  "conversationId": "2026701743246254107",
  "displayTextRange": [
    0,
    280
  ],
  "inReplyToUserId": null,
  "inReplyToUsername": null,
  "author": {
    "type": "user",
    "userName": "omarsar0",
    "url": "https://x.com/omarsar0",
    "twitterUrl": "https://twitter.com/omarsar0",
    "id": "3448284313",
    "name": "elvis",
    "isVerified": false,
    "isBlueVerified": true,
    "verifiedType": null,
    "profilePicture": "https://pbs.twimg.com/profile_images/939313677647282181/vZjFWtAn_normal.jpg",
    "coverPicture": "https://pbs.twimg.com/profile_banners/3448284313/1565974901",
    "description": "",
    "location": "DAIR.AI Academy",
    "followers": 291571,
    "following": 776,
    "status": "",
    "canDm": true,
    "canMediaTag": true,
    "createdAt": "Fri Sep 04 12:59:26 +0000 2015",
    "entities": {
      "description": {
        "urls": []
      },
      "url": {}
    },
    "fastFollowersCount": 0,
    "favouritesCount": 34909,
    "hasCustomTimelines": true,
    "isTranslator": true,
    "mediaCount": 4525,
    "statusesCount": 17379,
    "withheldInCountries": [],
    "affiliatesHighlightedLabel": {},
    "possiblySensitive": false,
    "pinnedTweetIds": [
      "2028103978190590118"
    ],
    "profile_bio": {
      "description": "Building @dair_ai • Prev: Meta AI, Elastic, PhD • New AI learning portal: https://t.co/1e8RZKs4uX",
      "entities": {
        "description": {
          "hashtags": [],
          "symbols": [],
          "urls": [
            {
              "display_url": "academy.dair.ai",
              "expanded_url": "https://academy.dair.ai/",
              "indices": [
                74,
                97
              ],
              "url": "https://t.co/1e8RZKs4uX"
            }
          ],
          "user_mentions": [
            {
              "id_str": "0",
              "indices": [
                9,
                17
              ],
              "name": "",
              "screen_name": "dair_ai"
            }
          ]
        },
        "url": {
          "urls": [
            {
              "display_url": "dair.ai",
              "expanded_url": "https://www.dair.ai/",
              "indices": [
                0,
                23
              ],
              "url": "https://t.co/XQto5ypSIk"
            }
          ]
        }
      }
    },
    "isAutomated": false,
    "automatedBy": null
  },
  "extendedEntities": {
    "media": [
      {
        "display_url": "pic.twitter.com/Wqh3qijgQz",
        "expanded_url": "https://twitter.com/omarsar0/status/2026701743246254107/photo/1",
        "ext_media_availability": {
          "status": "Available"
        },
        "features": {
          "large": {
            "faces": [
              {
                "h": 198,
                "w": 198,
                "x": 71,
                "y": 491
              }
            ]
          },
          "orig": {
            "faces": [
              {
                "h": 198,
                "w": 198,
                "x": 71,
                "y": 491
              }
            ]
          }
        },
        "id_str": "2026701740100431877",
        "indices": [
          281,
          304
        ],
        "media_key": "3_2026701740100431877",
        "media_results": {
          "id": "QXBpTWVkaWFSZXN1bHRzOgwAAQoAARwgSn6kWjAFCgACHCBKf1/boBsAAA==",
          "result": {
            "__typename": "ApiMedia",
            "id": "QXBpTWVkaWE6DAABCgABHCBKfqRaMAUKAAIcIEp/X9ugGwAA",
            "media_key": "3_2026701740100431877"
          }
        },
        "media_url_https": "https://pbs.twimg.com/media/HCBKfqRaMAUx88e.jpg",
        "original_info": {
          "focus_rects": [
            {
              "h": 762,
              "w": 1360,
              "x": 0,
              "y": 0
            },
            {
              "h": 1360,
              "w": 1360,
              "x": 0,
              "y": 0
            },
            {
              "h": 1550,
              "w": 1360,
              "x": 0,
              "y": 0
            },
            {
              "h": 1784,
              "w": 892,
              "x": 0,
              "y": 0
            },
            {
              "h": 1784,
              "w": 1360,
              "x": 0,
              "y": 0
            }
          ],
          "height": 1784,
          "width": 1360
        },
        "sizes": {
          "large": {
            "h": 1784,
            "w": 1360
          }
        },
        "type": "photo",
        "url": "https://t.co/Wqh3qijgQz"
      }
    ]
  },
  "card": null,
  "place": {},
  "entities": {
    "hashtags": [],
    "symbols": [],
    "urls": [
      {
        "display_url": "arxiv.org/abs/2602.17270",
        "expanded_url": "https://arxiv.org/abs/2602.17270",
        "indices": [
          927,
          950
        ],
        "url": "https://t.co/E1HCf9QzB4"
      }
    ],
    "user_mentions": []
  },
  "quoted_tweet": null,
  "retweeted_tweet": null,
  "isLimitedReply": false,
  "article": null
}