Source: raw/I_Tried_HeyGen_Video_1_Here_s_Everything_It_Can_Do.md — “I Tried HeyGen Video 1: Here’s Everything It Can Do” (youtube.com/watch?v=zcmczqIifCo, uploaded 2026-10-29). Two specs (clip length, resolution) come from raw/ChatGPT_Can_Make_Videos_Now._Here_s_the_Full_Workflow.md (youtube.com/watch?v=5Z7_7Ca5x-8), which uses the model inside ChatGPT. Studio-app specs (added 2026-11-01): ai-research/hyperframes-studio-app-video-clips.md. Creator: HeyGen official channel (@heygen_official); presenter introduces herself as “Mary Lou” (auto-caption spelling) URL: https://www.youtube.com/watch?v=zcmczqIifCo Duration: 4:51 Platform: YouTube Skill level: operator

HeyGen Video 1 is HeyGen’s own video-generation model: you describe a shot and it generates the picture and the sound together. It sits alongside HeyGen’s avatar products rather than replacing them. It makes short cinematic or stylized clips (B-roll, social cutaways), while a talking presenter is still the job of a personal avatar plus HeyGen Voice. This is a vendor-made walkthrough, so the ranking claim below is HeyGen’s own and was not checked here.

Key Takeaways

  • What it is. HeyGen’s in-house model. One prompt produces video and audio “all at once.” Selectable looks include live action, claymation, anime and screen print.
  • Vendor ranking claim. The presenter says Video 1 “ranked number one on Artificial Analysis text-to-video leaderboard.” The video gives no date, score or leaderboard snapshot; treat it as a vendor claim until checked against the leaderboard.
  • Three ways to use it: the HeyGen Video product page, Creative Studio inside the HeyGen app (longer clips, your own images), and an API for developers. It is also callable from the HeyGen plugin in ChatGPT; see HeyGen Plugin in ChatGPT.
  • Clip specs (ChatGPT source). “5 to 15 seconds, up to 2K” per clip, as stated in the ChatGPT workflow video.
  • Three input modes:
    • Text only.
    • A picture, which becomes the first frame of the clip.
    • References (a product photo, your actual shop). HeyGen says it keeps them in the scene.
  • Write the prompt like a shot list for a film crew, in four parts:
    1. Subject + one action: who or what is in the scene and the single thing happening. “Keep it to one thing.”
    2. Camera: how close it is and how it moves (slow push-in vs a locked-off shot).
    3. Look, mostly light: dark and moody vs light and sunny.
    4. Sound, and what to leave out: say what you want to hear, then exclusions such as “no music, no logo, no writing on the walls.” The presenter calls this the part most people skip.
  • Script Enhance rewrites a one-sentence idea into a full shot with camera, light and look. It does not add sound or exclusions, so add those yourself before generating. Edits can be undone or made by hand.
  • Settings to check before generating: length (leave on auto or pick seconds), quality, and shape (wide for YouTube, tall for Reels and TikTok).
  • What works best: one subject in one place doing one thing with a steady camera; objects being held or worn.
  • What HeyGen says is harder: close-up hand work; soft materials moving (hair, petals, paper); long on-screen text. Keep in-video words to a few, or add text in post.
  • Iterate one change at a time so you know which change fixed the shot.
  • Demo outputs shown: a coffee cup (one-sentence prompt + Script Enhance), a bakery at 4 a.m., a potter at the wheel, a claymation scene, an anime scene with bike and cicada sounds, and a screen-print “old movie poster.”
  • Where it fits: cutaways between talking segments so viewers are not watching the same shot throughout, and standalone short social clips. For a presenter talking for the whole video in your own voice, the presenter points to your personal avatar with HeyGen Voice (which she also calls “the number one voice model,” another vendor claim).
  • Prompt discipline overlaps HeyGen’s Seedance guidance. The four-part shot list follows the same logic as the subject / action / environment / camera / style structure taught for Avatar Shots + Seedance 2.0. The difference is that Video 1 adds an explicit sound-and-exclusions slot because it generates audio.

In the HyperFrames Studio App (added 2026-11-01)

Source: ai-research/hyperframes-studio-app-video-clips.md (HyperFrames docs: Make clips with HeyGen Video 1), first-party HeyGen documentation.

The HyperFrames Studio desktop app exposes Video 1 with different limits from the ChatGPT plugin specs above. The limits depend on the surface, so neither source contradicts the other:

SurfaceLengthShape and quality
ChatGPT plugin (per the ChatGPT video above)5–15 sUp to 2K
Studio app chat (# → HeyGen Video 1)5–60 s, default 1516:9, 9:16 or 1:1; 480p or 768p (default 768p)
Studio app Generate media form5–15 s16:9 or 9:16; a Quality setting
  • Steering in the Studio app: a start picture (exact first frame), up to 9 look-reference images, or one ≤5 s movement-reference video plus a picture of the performer. The movement option gives a single take of up to 15 s.
  • Cost in the Studio app: the HyperFrames plan if it covers it, otherwise HeyGen API credits. The chat path generates without a price check, and the form shows the price first. No per-second figure is published.
  • Full walkthrough: Studio App — Brand Kits, Reference Videos and Video 1 Clips.

Try It

  1. In HeyGen: home → Apps → Creative Studio → Video tab → choose HeyGen Video at the top.
  2. Write one prompt using the four slots: subject + one action, camera, light/look, sound + “no music, no logo, no on-screen text.”
  3. If you are stuck, type one sentence, click Script Enhance, then add the sound and exclusion lines it leaves out.
  4. Set length (auto), quality, and shape (tall for Reels/TikTok) before generating.
  5. For brand consistency, attach a product or location photo as a reference, or a still as the first frame.
  6. Change one variable per regeneration and note what improved.
  7. Drop the best clips between talking-head segments in an existing avatar video rather than building a whole video from Video 1 clips.

Implementation

Tool/Service: HeyGen Video 1 (HeyGen Creative Studio; HeyGen API; HeyGen plugin in ChatGPT) Setup: HeyGen account → Apps → Creative Studio → Video → HeyGen Video. Inputs: text, optional first-frame image, optional reference images. Cost: Not stated in either source. The ChatGPT source says the HeyGen free plan works with the plugin, but does not say whether Video 1 generations are included or how many credits they use. Integration notes: Output length is “auto” or user-picked. Aspect ratio is wide or tall. Per the ChatGPT source, clips run 5–15 seconds at up to 2K. For longer pieces, generate several clips and cut them together, or let Video Agent assemble them. The presenter says her prompts are in the YouTube description; they are not in the transcript.

Open Questions

  • Leaderboard claim: which Artificial Analysis leaderboard (text-to-video vs image-to-video), what date, and what score or Elo? Not given.
  • Pricing and credits per generation, and plan limits on quality or 2K output: not stated.
  • API surface: the endpoint name and parameters are not given. ^[inferred] It may correspond to the generic text-to-video / image-to-video tools on the HeyGen MCP connector, but that is unconfirmed.
  • Presenter name: “Mary Lou” is the auto-caption rendering and may be misspelled.
  • Prompts: the presenter says all prompts are in the video description; that text was not captured.
  • Max length in Creative Studio: the video says Creative Studio allows “longer clips” than the product page but gives no number. The 15-second figure comes from the ChatGPT plugin video.