@AlexM_SLC

Legal and BD lead @silencionetwork | Check out https://xitter.cf/t.co/wQBp9qFjeF

Joined January 2024
  • Tweets1K
  • Following307
  • Followers532
41Media
Exciting days ahead. Stay tuned 👀
55
39
21
265
18,776
“Most natural way to interact with intelligence” - if you speak a top 100 language. Could not agree more with the statement, but we do have a long way to go for the remaining 3.5B speakers globally This is why we built @silencioNetwork
jack
@jackOct 23
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence. block.xyz/dessa

Dessa - the Conversation Model trained only on speech

block.xyz
7
9
44
937
Last month a free speech corpus with more than a million hours, in 147 languages, went up for anyone to download. If your data strategy is "more hours", that's now free.
4
6
1
24
1,409
What isn't: a consent record for every speaker, accents labelled by the people who have them, and transcripts in the script the language is actually written in.
1
10
158
Long-tail languages used to be research projects at the big labs. In 2026 they're on commercial roadmaps. That changes the data question. A research demo can run on read speech. A product launch in Lagos or Cebu can't.
7
1
22
1,710
Most multilingual ASR benchmarks are read speech. Your users aren't reading. They talk, with accents, on ordinary phones, in noisy rooms. A model can top a read-speech leaderboard and still fail the call centre in Lagos. Evaluate on the speech you'll actually serve.
1
6
21
580
Asia-Pacific is growing conversational AI at 26.1% CAGR through 2030. The deepest commercial speech corpora still sit in English, French, German, Spanish. The gap between where the buyers are and where the data lives is the whole commercial thesis for specialist speech data.
1
4
1
26
428
The buyer momentum is in Southeast Asia, South Asia, MENA. The corpora that meet enterprise procurement bars are in Europe and North America. Three years of arbitrage until that gap closes. Whoever ships auditable data for the actual buyer languages wins.
1
4
114
This is the operational shape of the market I keep repeating in deal calls. Buyer languages are not the same as available languages. That is a 3-year fix, not a 3-month fix. Source: Grand View Research, Fortune Business Insights, 2026. #SpeechAI #VoiceAI #MultilingualAI
4
53
Dear Silencians, please help us with our Wolof data collection and tag any crypto X communities from Senegal! @silencioNetwork OR share the ongoing Wolof project with these communities, using your referral code, and you can earn referral commissions!!
3
8
1
43
1,780
Back from Interspeech 2026, Sydney. Every major voice AI lab in the room. The papers that mattered most: the ones on languages public models still can't hear. Silencio was Full-Day Sponsor of Indigenous Voices Day. That was the point of showing up.
6
7
1
43
1,017
Final day at Interspeech 2026. Best week of speech AI in a year, easy. If you build in this space and were not here, put next year on the calendar. Booth S5 open one more day.
1
5
23
650
Day 3 at Interspeech, Sydney. Three days of conversations with people who take low-resource speech data seriously and know how hard it actually is. Not many places you can find this many of them in one room.
3
2
28
708
Today is Indigenous Voices Day. Silencio is Full-Day Sponsor. The reason to be here is the room. Researchers, elders, engineers, people who spent their careers on languages the internet forgot. Not many days in this job feel more like the point of it.
2
6
37
601
Day 1 at Interspeech, Sydney. The corridor conversations already tell you the whole story. Everyone here is thinking about the same problem: the languages public data does not cover. Come find us at Booth S5.
2
4
37
732
Made it to Sydney for Interspeech 2026. The Silencio team is in. Booth S5 opens tomorrow. Full-Day Sponsor of Indigenous Voices Day on Tuesday. Wide awake, watching the Opera House do its thing. Ready for the week.
4
6
36
682
Heading to Interspeech 2026 in Sydney this week. Every major speech AI lab in one room. The day worth watching: Tuesday, Indigenous Voices Day. Silencio is Full-Day Sponsor. We bring consented, real-world data for the 6,000 languages the internet has never held.
4
8
2
41
1,747
Public service announcement. Record a few clips for @silencioNetwork and you get two things: 1. Voice AI that actually hears your language. 2. A private safeword only your model recognises, for when the robots take over. Yours will listen. Everyone else's will not.
9
9
1
50
1,520
Every major AI lab will need a voice-data partner or an acquisition within 36 months. The reason is structural. In-house collection cannot reach the tail. Purchased data with the wrong provenance is a legal problem. Voice data companies sit where CV data companies did in 2016.
2
8
1
39
2,305
Mozilla Common Voice v18: ~31,841 hours across 129 languages. That is the largest open, commercial-cleared multilingual speech dataset in the world. Modern frontier speech models train on multiples of that.
2
9
25
624
Which means everything above the Common Voice ceiling is not open, not free, and, more often than not, not documented. That gap is the real training-data question in 2026. What is inside the training set, where did it come from, and can the provider prove it.
1
5
154
For serious multilingual voice AI, the audit trail matters more than the hour count. Silencio builds on the auditable side. Common Voice is the floor. The industry needs the ceiling raised by 10x, provenance intact. Source: Mozilla Common Voice 18. #SpeechAI #MultilingualAI
1
5
116
Load more