HypeReel

Why a clip feature ships in a coaching product

Real-time coaching helps you get to highlight-worthy moments. Without a way to share them, those moments live and die in your local recordings folder. HypeReel closes that loop. The same persona voices that coach you live — Nova, Luna, Kaze, Aura, Mika, Ren on the 3D avatar side — are available in the HypeReel voice picker, so the clip you post the next morning can land in a voice that already means something to you.

The continuity matters more than it sounds. A clip narrated by a generic AI voice feels like a tool. A clip narrated by a persona voice you recognize — whether it's the same companion that was in your headset during the fight, or a different persona that fits the clip's audience — feels like a story worth telling. Players who use both halves of the product end up treating Sidekick differently — less “helper app,” more “set of companions who show up.”

How HypeReel beats manual editing

Manual highlight editing has three slow steps: scrub raw footage to find the moment, trim around it, write or record commentary. HypeReel collapses two of those into one automated pass and removes the third. Vision AIscans the footage for notable moments and timecodes them. The narration layer generates voice commentary using your companion's tone, tuned for the kind of moment it just saw, and lines the spoken beats up with the highlights. The manual scrub-then-narrate cycle goes away.

The output isn't always perfect. Sometimes the AI picks the wrong moment to hype, sometimes the narration is dry. The product is designed for the case where you want a 90%-good clip in 2 minutes instead of a 100%-good clip in 30 minutes. Most creators will take that trade.

How the highlight pipeline actually works

The HypeReel pipeline has four steps: ingest, detect, trim, narrate-and-render. Each is what it sounds like, but the way they connect is what makes the output feel intentional instead of templated.

Ingest is the raw footage you give it — a recording, a screen capture, a Twitch VOD segment. HypeReel accepts standard video formats and pulls frames from the uploaded video so the pipeline has something to score.

Detectis where vision AI finds the notable moments — boss kills, clutch dodges, rare drops, satisfying combos. The detector looks at what the screen actually shows, not audio peaks or hype-meter heuristics. A three-minute clip might have one big moment or four. The pipeline marks all of them with timecodes. This layer is what general-purpose video editors (CapCut, Veed.io) don't ship — it's game-aware moment recognition, not waveform analysis.

Trim identifies the tight window around each detected moment — timecodes that mark where the highlight payoff lands so the narration can be tuned to those beats.

Narrate-and-render generates a script tuned to what the detector saw (the kind of moment, the game it's in, the player state at the moment) and renders that script in the voice you picked at clip time. The final render bundles the source video and the narration audio together so the spoken beats line up with the highlights on screen.

Pick the voice for the clip

HypeReel ships a voice picker at clip time. You choose the voice the narration uses — including the same voices the Sidekick personas speak in during live coaching. So a Souls retro clip can land in Nova's measured, analytical voice; an Apex highlight can land in Luna's louder, more celebratory voice; the same kill clip narrated by Mika sounds warm and patient instead. Different clip, different voice, same workflow.

Why this is a clip-time choice rather than auto-inheriting your live companion: a player's coaching voice and their content voice aren't always the same. Someone who runs Nova as a coach because the analytical tone helps during boss attempts may want Luna narrating the boss-kill clip because the celebratory tone fits the shareable moment. HypeReel doesn't lock those choices together. The voice picker exposes the persona voices alongside any other voices the system supports, so you can match the clip to the audience it's going to.

The persona voices themselves are the same six described on /features/3d-avatar: Nova(calm & strategic), Luna(hype & energetic), Kaze(cool & mysterious), Aura(sharp & precise), Mika(gentle & supportive), and Ren(bold & fearless). Browse the full cast at the Avatar Gallery.

Why competitors haven't shipped this

The AI gaming companion category is young. The product axes most companies are competing on are chat (Character.AI, Replika), persona variety (Questie.ai), or hardware ecosystem (Razer Ava). None of them have shipped an AI-narrated highlight-clip pipeline. The gap is real and worth thinking about.

Two reasons it hasn't happened. First, narrated-clip output requires a coherent persona system — generic AI voice on a generic clip feels worse than a player's own narration in their own voice. Sidekick can ship HypeReel because the same six personas that coach you live are available to narrate your highlights; products that ship chat-character marketplaces or roleplay companions don't have one coherent persona to put on a clip.

Second, the clip workflow is downstream of game-awareness. A clip narrator has to know what's worth hyping in the footage — what's a clutch dodge versus a fluky frame, what's a rare drop versus a normal one, what's a phase transition worth calling out. Sidekick's vision layer was already doing that work for live coaching, so HypeReel rides on top of it. Building HypeReel without the upstream vision investment would mean training a separate clip-moment detector — a bigger lift than most product roadmaps absorb.

So HypeReel doesn't have a direct competitor today. The closest workflow is “raw clip plus your own narration in CapCut or Veed.io,” which is exactly the 20-minute manual workflow the product is built to replace.

Creator versus casual use

HypeReel is one of those features that means different things to different players. For creators, it's a content pipeline — every notable session becomes a piece of social content with consistent branding. For casual players, it's a way to share cool moments with friends without learning a video editor. Both flows ship in the same product, gated by how many clips you run through it. Browse the Reels Gallery for example outputs.