Twitch sentiment analysis

Twitch sentiment analysis, in real time.

Uncloak scores every Twitch chat message the moment it arrives. Words and emotes, in 35 languages, turned into a live read on how an audience is reacting.

Live chatStored score per message
Gamescom Opening Night Live2 min intervals
4,003,833chat messages scored across 1,950 channels
Average sentiment per 2 min, 18:00 to 21:00 UTC
21.5%Positive
65.1%Neutral
13.5%Negative

Real messages from Gamescom Opening Night Live 2026 chat on 25 August 2026, with their stored scores. Usernames removed.

~1,000chat messages processed per second
Instanteach message is scored as it arrives
−1 to +1a sentiment score for every message
35languages supported for sentiment
~2 minrefresh for stream sentiment and engagement
Real event

Gamescom Opening Night Live, read live across every stream.

On 25 August 2026, the show ran on official channels and hundreds of co-streams at once. Uncloak scored every message as it arrived and summarised all streams together every two minutes.

1,950channels with chat
4,003,833chat messages scored
286,326unique chatters
82,964messages in the busiest 2 minutes
18:12gears · war · xbox19.0%19:06max · mad · aion21.0%19:20guitar · hero26.8%19:58witcher · dlc27.8%

Hover the chart for each two minute interval. Click the legend to show or hide a series.

Source: Uncloak, Gamescom Opening Night Live, 25 August 2026, 18:00 to 21:00 UTC, all Twitch streams covering the event. Share of all messages in each two minute interval; the rest were neutral. Labels show the words chat used most in that interval.

Text and emotes

On Twitch, the emote often carries the meaning.

Uncloak reads both: the text with VADER based NLP, and the emotes with a custom mapping of thousands of Twitch emotes, each one classified by our founder. “Sub sunday game Pog” has no sentiment words at all. The emote makes it positive.

With emotes · Positive
Sub sunday game Pog+0.72
KEKW actually good+0.85
With emotes · Negative
Everyone on their phone at live event NotLikeThis−0.54
Aware its joever−0.67
Text only · Positive
i play it for 2 years its so good+0.73
There are so many great stories that haven't been told yet+0.66
Text only · Negative
animations are really bad−0.58
DnD but spells just slop, where warlock spells? WTF?−0.83

Real messages from Gamescom Opening Night Live 2026 chat on 25 August 2026, with their stored scores. Usernames removed.

How it works

From chat message to stream summary.

  1. 1

    Every message, as it lands

    Uncloak processes around 1,000 chat messages per second, across 35 languages.

  2. 2

    Text and emotes, scored together

    The moment a message arrives, its words and emotes are analysed and it gets a score between −1 and +1.

  3. 3

    A live read on the stream

    Stream sentiment and engagement metrics refresh about every two minutes, so you see how the room is moving.

Messages are analysed immediately. Two minutes is the refresh interval for stream level metrics.

Emote mapping

Thousands of Twitch emotes, classified by hand.

Twitch chat has its own language. Our founder personally classified thousands of emotes, so a LETSGO and a Sadge count for what chat means by them.

Hype
  • LETSGO
  • POGGERS
  • catJAM
  • WICKED
  • peepoClap
Laughter
  • ICANT
  • PepeLaugh
  • xdd
  • catKISS
  • ratJAM
Disbelief
  • AINTNOWAY
  • WHAT
  • HUH
  • D:
  • modCheck
Letdown
  • NOOOO
  • Sadge
  • SadgeCry
  • TrollDespair
  • Clueless
Tension
  • monkaW
  • monkaSTEER
  • RAGEY
  • Madge
  • peepoRiot
Hope and respect
  • Prayge
  • COPIUM
  • o7
  • NODDERS
  • Chatting

Popular emotes from 7TV, grouped by how chat typically uses them. Emote images via 7TV.

Use cases

Built for moments that happen live.

Go beyond positive and negative.

Sentiment tells you which way a reaction leans. Uncloak AI goes further, into emotions such as hype, fear or anxiety, and what triggered them.

LETSGOHypemonkaWFearmonkaSTEERAnxietyRAGEYFrustrationICANTJoySadgeDisappointment
Explore emotion analysis
FAQ

Questions about Twitch sentiment analysis

It measures whether Twitch chat reacts positively, negatively or neutrally to a stream, a brand, a game or a moment. Uncloak does it live, message by message, and rolls it up into stream level metrics.

Each message is analysed the moment it arrives. The text is read with VADER based NLP and the emotes with a custom mapping of thousands of Twitch emotes. Every message gets a sentiment score between −1 and +1.

Yes. Our founder personally classified thousands of Twitch emotes, so emotes count toward the score alongside the words.

Sentiment analysis supports 35 languages, so chat is scored across international streams, not just English ones.

Uncloak processes around 1,000 chat messages per second, and each message is scored as soon as it arrives.

Stream sentiment and engagement metrics refresh approximately every two minutes. Individual messages are analysed immediately; two minutes is the refresh interval for stream level metrics.

See how your audience really reacts.

Get a walkthrough of live Twitch sentiment on a stream, brand or game you care about.