Source: Discord channel 1483773846338142229 (guild 1482384783739715694), last 78 messages (Apr 7 → ~Apr 28 2026), exported with @mdp/discord-dump.
Koe (koe.sh) is an AI video/trailer generation platform — a "video production engine" that chains models (Kling, LTX 2.3 Pro, Seedance 2.0) into a storyboarded workflow: custom characters, scene editor, camera angles, duration/motion, preview frames, voice cloning, lip sync, and templates (Gothic 90s Anime, Pixar 3D Animated). It's run by jordyx (founder, based in NY), and the Discord is the community/support channel.
Very small and very early — only ~5 real participants across 78 messages:
- jordyx — the founder, who posts the bulk of the messages (launch announcements, support replies, hype posts like "Our growth rate is exponentially increasing everyday!! 😇").
- vitamins — an early, friendly tester.
- DIA — an enthusiastic paying user ($9/mo plan) who calls it "the best AI trailer gen on the market."
- Erin and Beloved — two skeptical would-be users who raised trust/pricing concerns.
- Arkki — the most thoughtful user; a media-art MA who did a long comparative review vs. a competitor ("Freebeat").
So this isn't a broad community consensus — it's a handful of early users plus the founder. That caveat matters a lot.
There's no single unified consensus, but the signals cluster into three groups:
- DIA: "you have the best ai trailer generator on market" (said twice), and was so impressed he connected the founder to a film-financing contact (Darren Aronofsky-adjacent). He's a paid user.
- Arkki: a long, detailed review concluding Koe "just clicks" for him and has a "soulful and personal" feel competitors lack. He explicitly differentiated Koe from Freebeat: Koe is for people who want hands-on creative control (timeline editor, iterative workflow), while Freebeat is a "leave it to autopilot" service. He said he'd pay even without the video generation, just for the ideation/storyboarding tool.
- Both framed the differentiator as creative control / "human agency," not raw output quality. Arkki even urged the founder not to automate more: "pls don't push for it to automate more... the intuitive timeline tool + koe agent is gold, that sets it apart."
- Erin asked for the Terms of Service before signing up: "I want to make sure me and my content are protected." The founder had to point her to a footer link he'd "newly added" — i.e., ToS was missing until recently.
- Beloved's complaint is the sharpest in the whole dump: the marketing implies a free trial, but after uploading content and waiting, he hit a paywall. His read: "So is there an actual free tier or not because it doesn't look like it in practice." And on the pricing page: "your copy on the pricing page for the free tier is misleading... what exactly am I meant to be trying out?"
- The founder's honest reply: "Currently, we can't give out free generations because we will be losing money." His fallback offer was "try the $12 plan, I'll refund you if you don't like it."
- DIA also hit a billing bug — couldn't upgrade from $9 to $25/mo ("it goes to the same screen"), and on the $9 plan couldn't generate a single video despite paying.
- A bit of off-topic (a $40–50 gig request from Ritwik, a TikTok link, Suno music-video tutorials, referral program announcements). Mostly founder-driven announcements rather than community discussion.
If I had to compress it to a sentence:
The few real users think the underlying product is genuinely good — arguably best-in-class for people who want hands-on creative control over AI video — but the go-to-market has real trust problems: a misleading "free trial," billing bugs, and a until-recently-missing Terms of Service.
More specifically:
- Product quality / differentiation: positive. The supporters' praise is specific and credible (creative control, iterative timeline, "soulful" UX) rather than generic hype. The competitive moat they identify is editorial control, not model quality — jordyx himself says "the models are not as important," the engine/architecture is.
- Trust / transparency: negative, and it's the recurring friction. Two of the ~5 real users raised it independently (missing ToS, misleading free-tier copy, billing bugs blocking even paid generation). The founder is responsive and fixes things, but they're reactive fixes, not airtight UX.
- Size/stage: extremely early. "Exponential growth" messaging from the founder is not corroborated by channel activity — it's a handful of users in a quiet channel over ~2.5 weeks (Apr 7 → late April/early May 2026).
- Sample size is tiny — 78 messages, ~5 real voices, one of whom is the founder. "Consensus" overstates it; this is closer to "early-user read."
- Channel bias — a support/community channel self-selects for people who care enough to show up; happy users and angry users are overrepresented, silent majority is invisible.
- Founder-heavy — a large fraction of messages are jordyx announcing features (storyboard mode, Seedance 2.0, lip sync, templates, forum). That's marketing, not community sentiment, so I weighted it lightly.
- Date window — only the latest 78 messages (Apr 7 → ~Apr 28 2026). Older history or other channels could tell a different story.