@elonmusk
AI will do this and more
Small improvements to Imagine continue; Imagine 2.0 needs few weeks training for speech/audio and face consistency upgrades. Support 62.20%, Confront 12.20%
Small improvements to Imagine are happening frequently. Looks like we need another few weeks of training for Imagine 2.0, which will have major upgrades in speech/audio and face/details consistency.
Real-time analysis of public opinion and engagement
What the community is saying — both sides
many replies celebrate Imagine’s trajectory — “2.0 will break the internet,” creators are excited to make cinematic, viral content, and users call Grok Imagine a game‑changer.
the single most repeated demand is reliable facial identity, lip sync, and natural audio; fix those and AI video becomes professionally usable.
users interpret frequent incremental updates as a smart dev strategy — daily tweaks now, then a larger Imagine 2.0 jump that’s worth waiting for.
people want one‑person studios, educational videos, short films, translated/dubbed content, personalized Tesla news channels, and even first‑person driving simulations.
voice prompts and native/controllable voices, likeness/cameo support, saved character presets, an editing studio (CapCut‑style), LoRA support, longer/4K clips, and better multiscene control.
common complaints include wonky faces, robotic voices, bad hands/fingers, broken infographics/text, and latency issues that hinder testing and workflows.
at least one technical suggestion calls for an early NG keyword check to disable “boost” on risky prompts — reducing runaway outputs, compute cost, and moderation failures.
users say the community guide and shared prompts already dramatically improve results and ask for more tutorials, early access testing, and tooling around prompts.
creators and investors see a large 2026 opportunity — scalable multimodal tools could upend content production and creator economics.
numerous replies thank the team, share successful outputs made with Grok, and express loyalty — many are already evangelists who’ve replaced paid workflows.
heavy floating particulate noise ("crushed‑leaf/glitter"), leaf‑blower particle clouds, unintended scene jumps, poor temporal memory, identity drift/character morphing, wardrobe hallucinations and weak prompt adherence.
occasional cleaner builds exist, but releases often fluctuate in quality and some users say today's outputs are notably worse than a week ago.
reference loss after the first splice, hallucinations on extensions and steep quality degradation on longer clips, rendering extended outputs unusable.
(promised R‑rated ability feels G‑rated), while others demand stricter enforcement against illegal content (ban CP). The theme is inconsistent, "bipolar" moderation behavior.
and lack of pro features — all contributing to the sense the app is a "fun toy" rather than a professional tool.
(some say ChatGPT makes better images), and warn that if temporal/artifact problems persist, competitors will overtake.
"more training, less promises" and fewer public predictions — users want demonstrated improvements, not roadmaps or proclamations.
(e.g., put energy into Tesla/fewer distractions), reflecting frustration with perceived mission drift.
Most popular replies, ranked by engagement
AI will do this and more
w kind of 24-hour-a-day channel. One on any topic. Your favorite sports team. The AI news of the day. The war in Iran. Keeping up with Elon. Etc etc etc. Let Grok really study lists in great detail. Write a script. Shove it over to Imagine. Build a show. Or a piece therein.
Grok Imagine
What about censorship? Because the censorship, even for those who know to dance around such things, is getting to be a pain.
This is 2026 and Afghan women situation💔
Small improvements on the moon
Found something wrong with this article? Let us know and we'll look into it.