vidtheque.

read-only demo

Videos SS-A8sE7hkw

Sovereign Escape Velocity: Ownership w Open Models — Gus Martins, & Ian Ballantyne, Google DeepMind

index_state ready data_status ok

AI Engineer· published 2026-06-10· 0:20:52· en-US· indexed 2026-08-11 10:49

Open on YouTube

Scene timeline

  1. Shot 0, 0:00 to 0:05, 1 of 1 keyframes kept
  2. Shot 1, 0:05 to 0:09, 1 of 1 keyframes kept
  3. Shot 2, 0:09 to 0:14, 1 of 1 keyframes kept
  4. Shot 3, 0:14 to 0:27, 1 of 1 keyframes kept
  5. Shot 4, 0:27 to 0:43, 1 of 1 keyframes kept
  6. Shot 5, 0:43 to 0:45, 1 of 1 keyframes kept
  7. Shot 6, 0:45 to 1:00, 1 of 1 keyframes kept
  8. Shot 7, 1:00 to 1:10, 1 of 1 keyframes kept
  9. Shot 8, 1:10 to 1:13, 0 of 1 keyframes kept
  10. Shot 9, 1:13 to 1:15, 0 of 1 keyframes kept
  11. Shot 10, 1:15 to 1:18, 0 of 1 keyframes kept
  12. Shot 11, 1:18 to 1:32, 0 of 1 keyframes kept
  13. Shot 12, 1:32 to 1:45, 1 of 1 keyframes kept
  14. Shot 13, 1:45 to 2:00, 0 of 1 keyframes kept
  15. Shot 14, 2:00 to 2:15, 0 of 1 keyframes kept
  16. Shot 15, 2:15 to 2:21, 0 of 1 keyframes kept
  17. Shot 16, 2:21 to 2:48, 1 of 1 keyframes kept
  18. Shot 17, 2:48 to 3:14, 1 of 1 keyframes kept
  19. Shot 18, 3:14 to 3:41, 0 of 1 keyframes kept
  20. Shot 19, 3:41 to 4:20, 1 of 1 keyframes kept
  21. Shot 20, 4:20 to 5:02, 1 of 1 keyframes kept
  22. Shot 21, 5:02 to 5:30, 0 of 1 keyframes kept
  23. Shot 22, 5:30 to 5:58, 0 of 1 keyframes kept
  24. Shot 23, 5:58 to 6:31, 1 of 1 keyframes kept
  25. Shot 24, 6:31 to 7:00, 1 of 1 keyframes kept
  26. Shot 25, 7:00 to 7:32, 1 of 1 keyframes kept
  27. Shot 26, 7:32 to 7:58, 0 of 1 keyframes kept
  28. Shot 27, 7:58 to 8:23, 1 of 1 keyframes kept
  29. Shot 28, 8:23 to 8:58, 1 of 1 keyframes kept
  30. Shot 29, 8:58 to 9:33, 0 of 1 keyframes kept
  31. Shot 30, 9:33 to 10:04, 1 of 1 keyframes kept
  32. Shot 31, 10:04 to 10:34, 0 of 1 keyframes kept
  33. Shot 32, 10:34 to 11:05, 1 of 1 keyframes kept
  34. Shot 33, 11:05 to 11:36, 0 of 1 keyframes kept
  35. Shot 34, 11:36 to 12:06, 1 of 1 keyframes kept
  36. Shot 35, 12:06 to 12:36, 0 of 1 keyframes kept
  37. Shot 36, 12:36 to 13:02, 1 of 1 keyframes kept
  38. Shot 37, 13:02 to 13:29, 0 of 1 keyframes kept
  39. Shot 38, 13:29 to 13:55, 0 of 1 keyframes kept
  40. Shot 39, 13:55 to 14:22, 1 of 1 keyframes kept
  41. Shot 40, 14:22 to 14:48, 0 of 1 keyframes kept
  42. Shot 41, 14:48 to 15:25, 1 of 1 keyframes kept
  43. Shot 42, 15:25 to 16:02, 1 of 1 keyframes kept
  44. Shot 43, 16:02 to 16:04, 1 of 1 keyframes kept
  45. Shot 44, 16:04 to 16:11, 1 of 1 keyframes kept
  46. Shot 45, 16:11 to 16:21, 1 of 1 keyframes kept
  47. Shot 46, 16:21 to 16:23, 0 of 1 keyframes kept
  48. Shot 47, 16:23 to 16:48, 1 of 1 keyframes kept
  49. Shot 48, 16:48 to 16:52, 0 of 1 keyframes kept
  50. Shot 49, 16:52 to 17:15, 1 of 1 keyframes kept
  51. Shot 50, 17:15 to 17:18, 0 of 1 keyframes kept
  52. Shot 51, 17:18 to 18:04, 1 of 1 keyframes kept
  53. Shot 52, 18:04 to 18:14, 1 of 1 keyframes kept
  54. Shot 53, 18:14 to 18:17, 0 of 1 keyframes kept
  55. Shot 54, 18:17 to 18:19, 1 of 1 keyframes kept
  56. Shot 55, 18:19 to 18:46, 1 of 1 keyframes kept
  57. Shot 56, 18:46 to 19:14, 1 of 1 keyframes kept
  58. Shot 57, 19:14 to 19:41, 0 of 1 keyframes kept
  59. Shot 58, 19:41 to 20:09, 0 of 1 keyframes kept
  60. Shot 59, 20:09 to 20:16, 1 of 1 keyframes kept
  61. Shot 60, 20:16 to 20:37, 1 of 1 keyframes kept
  62. Shot 61, 20:37 to 20:51, 1 of 1 keyframes kept

62 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
276
whisperx 276
chunks
37
from 276 cues
keyframes
38
kept of 62 captured
frames with text
38
1,239 lines read
chapters
0
from the source metadata
keyframe bytes
5.9 MB
word timings on 276 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-11 10:45 1m 08s
stt done — 2026-08-11 10:46 23s
chunk done — 2026-08-11 10:47 0s
text_embed done — 2026-08-11 10:47 1s
keyframe done — 2026-08-11 10:47 1m 47s
ocr done — 2026-08-11 10:49 20s
frame_embed done — 2026-08-11 10:49 6s

Frames, and what the machine read

  • 0:03 #0 done2 line(s)

    shot 0·sharpness 658.4

    1. AlEngineer0.98
    2. EUROPE1.00

  • 0:08 #1 done2 line(s)

    shot 1·sharpness 829.1

    1. PRESENTINGSPONSOR1.00
    2. Google DeepMind1.00

  • 0:13 #2 done3 line(s)

    shot 2·sharpness 907.6

    1. PLATINUM SPONSORS0.98
    2. # Braintrust0.96
    3. WorkOS OpenAI0.95

  • 0:16 #3 done8 line(s)

    shot 3·sharpness 312.7

    1. Sovereign1.00
    2. Gemma41.00
    3. Escape1.00
    4. Velocity1.00
    5. Al Ownership with0.98
    6. Open Models1.00
    7. Gemma0.81
    8. AlEng0.95

  • 0:31 #4 done7 line(s)

    shot 4·sharpness 300.1

    1. Sovereign1.00
    2. Gemma 40.94
    3. Escape1.00
    4. Velocity1.00
    5. Al Ownership with0.97
    6. Open Models1.00
    7. AlEngi0.90

  • 0:44 #5 done12 line(s)

    shot 5·sharpness 1402.1

    1. ★0.99
    2. AIE1.00
    3. ★1.00
    4. ★1.00
    5. Gus Martins1.00
    6. lan Ballantyne0.99
    7. Product Manager1.00
    8. Developer Relations Engineer1.00
    9. Google DeepMind1.00
    10. Google DeepMind1.00
    11. Google DeepMind0.98
    12. AIEnai0.84

  • 0:54 #6 done7 line(s)

    shot 6·sharpness 946.5

    1. ★1.00
    2. AIE1.00
    3. Gemma 40.99
    4. ★1.00
    5. ★1.00
    6. Google DeepMind0.98
    7. AlEngii0.94

  • 1:02 #7 done6 line(s)

    shot 7·sharpness 938.8

    1. ★0.83
    2. AIE1.00
    3. ★1.00
    4. Gemma 40.95
    5. ★1.00
    6. Engineering the future of Al1.00

  • 1:10 #8 skipped

    shot 8·duplicate of #7

  • 1:14 #9 skipped

    shot 9·duplicate of #6

  • 1:16 #10 skipped

    shot 10·duplicate of #7

  • 1:20 #11 skipped

    shot 11·duplicate of #7

  • 1:44 #12 done8 line(s)

    shot 12·sharpness 927.2

    1. ★1.00
    2. AIE1.00
    3. Gemma 40.99
    4. ★1.00
    5. ★1.00
    6. AlEngineer0.97
    7. EUROPE1.00
    8. IEn0.86

  • 1:47 #13 skipped

    shot 13·duplicate of #12

  • 2:02 #14 skipped

    shot 14·duplicate of #12

  • 2:17 #15 skipped

    shot 15·duplicate of #12

  • 2:24 #16 done8 line(s)

    shot 16·sharpness 1220.4

    1. Effective 2B1.00
    2. Effective 4B0.99
    3. AIE1.00
    4. ★1.00
    5. ★1.00
    6. AlEngineer0.97
    7. EUROPE1.00
    8. AlEnc0.82

  • 3:06 #17 done6 line(s)

    shot 17·sharpness 1226.3

    1. Effective 2B1.00
    2. Effective 4B0.96
    3. ★0.90
    4. AIE1.00
    5. ★1.00
    6. Engineering the future of Al1.00

  • 3:20 #18 skipped

    shot 18·duplicate of #17

  • 4:01 #19 done9 line(s)

    shot 19·sharpness 1248.2

    1. 26B A4B0.95
    2. 31B Dense0.97
    3. ★1.00
    4. AIE1.00
    5. ★1.00
    6. ★1.00
    7. AlEngineer0.96
    8. EUROPE1.00
    9. AlEngi0.94

  • 4:41 #20 done26 line(s)

    shot 20·sharpness 1152.4

    1. Gemma41.00
    2. Arena Elo Score1.00
    3. AIE1.00
    4. ★1.00
    5. 14521.00
    6. 14561.00
    7. 14541.00
    8. 14501.00
    9. ★1.00
    10. ★1.00
    11. ★1.00
    12. 14411.00
    13. 14251.00
    14. Gemma 40.98
    15. Glm 50.99
    16. Kimik2.50.99
    17. Qwen 3.50.99
    18. Deepseek v3.20.98
    19. 31B1.00
    20. 26B1.00
    21. 754B1.00
    22. 1100B1.00
    23. 397B1.00
    24. 685B1.00
    25. Engineering the future of Al0.99
    26. AlEngir0.96

  • 5:16 #21 skipped

    shot 21·duplicate of #17

  • 5:33 #22 skipped

    shot 22·duplicate of #17

  • 6:11 #23 done34 line(s)

    shot 23·sharpness 1060.2

    1. ai.dev1.00
    2. Playground1.00
    3. a0.60
    4. <> Get code0.97
    5. Gemma 4 318 IT0.94
    6. Explore Google models1.00
    7. ★1.00
    8. System instructions1.00
    9. AIE1.00
    10. Featured1.00
    11. Code and Chat1.00
    12. Image Generation1.00
    13. ★1.00
    14. No API Key0.95
    15. ★1.00
    16. ★1.00
    17. ★1.00
    18. Video Generation0.99
    19. Speech and Music1.00
    20. Real-time1.00
    21. Temperature1.00
    22. Media resolution0.99
    23. Default0.99
    24. Start building →0.98
    25. Thinking level1.00
    26. High1.00
    27. Structured outputs1.00
    28. Tools0.96
    29. G Grounding with Google Search0.99
    30. Code execution0.97
    31. Function callina1.00
    32. AlEngineer0.97
    33. EUROPE1.00
    34. AlEngir0.96

Transcript

276 cues· 3,827 words· 19,723 chars

  1. 0:15 Hi, everyone.
  2. 0:16 Can you hear me?
  3. 0:17 Yes, you can hear me.
  4. 0:18 Hi.
  5. 0:18 Sorry, sorry.
  6. 0:19 One minute late.
  7. 0:20 I'll try to do my best to finish earlier so my friend can do pretty cool demos for you.
  8. 0:25 I'm Gus.
  9. 0:26 This is Ian.
  10. 0:27 We are from the Google DeepMind.
  11. 0:30 And I'm specifically working on the Gemma product.
  12. 0:32 Do any of you know what Gemma models are?
  13. 0:36 OK, perfect, perfect.
  14. 0:37 Thank you very much.
  15. 0:39 So today, we are going to talk a little bit about ownership and open models.
  16. 0:44 And well, you know who we are.
  17. 0:45 But the idea is, last Thursday, we released our new family of models, Gemma 4.
  18. 0:51 And I'm going to talk a little bit about them.
  19. 0:53 There is going to be more information tomorrow in the keynote by Omar.
  20. 0:57 And there is another talk by Cassidy also tomorrow that she will go into even more details.
  21. 1:02 We are going to tell a little bit of the story, but the story is a little bit bigger.
  22. 1:06 We'll try our best here.
  23. 1:09 So why does it matter?
  24. 1:11 If you ask me, I work for Google, of course, if you ask me what's the best model for you to try, the easiest one I will answer for you, Gemini.
  25. 1:17 Gemini is the best model we have.
  26. 1:19 Pretty strong, multi-model, can do all kinds of things.
  27. 1:24 But then there's more to the story than just having the strongest model possible.
  28. 1:30 In some situations, you want to own the model.
  29. 1:35 You want to be able to run on your own hardware.
  30. 1:38 You want to customize it.
  31. 1:40 You want to be able to send your proprietary data that cannot leave your infrastructure.
  32. 1:45 So there are many situations where even the best proprietary model will not be able to help you directly.
  33. 1:52 That's when you might need an open model.
  34. 1:54 That's where Gemma comes in.
  35. 1:55 So when you think, why does Google have two family of models?
  36. 1:59 Because they complement each other.
  37. 2:01 So Gemma is the most intelligent one, can do a lot of cool stuff, but it's hosted in Google servers.
  38. 2:07 You need the API to access.
  39. 2:09 If you need more control and access, you need an open model.
  40. 2:13 That's why we have Gemma.
  41. 2:14 That's why, and we are very proud that the quality is very, very strong.
  42. 2:18 We're going to go through some details later.
  43. 2:19 But the idea is you would be able to do a lot of cool stuff with it.
  44. 2:23 Among the launches, we released four sizes.
  45. 2:26 Two are target to mobile, let's say like that, or IoT, or smaller devices.
  46. 2:33 It's an E2B and an E4B.
  47. 2:36 These names is a little bit weird.
  48. 2:37 We are the only ones that use this name.
  49. 2:39 And the E stands for effective.
  50. 2:42 And the idea here is...

Open at this second