Claude + Higgsfield / heygen + Blotato: Faceless Reels on Autopilot
NBA Anime Reel Cowork Prompt
Create a ~21-second vertical (9:16) anime short reel using the Higgsfield connector, fully self-serve, called "Hardwood Legends — The Takeover."
Story (5 chained shots): In an empty indoor gym (no crowd), the 2026 NBA Finals rivals face off — Jalen Brunson (Knicks, white "NEW YORK" #11 jersey, cornrow braids, full beard, stocky) vs Victor Wembanyama (Spurs, black "SPURS" #1 jersey, clean-shaven, slender giant). (1) Tense stare-down, Wemby leaning over Brunson; (2) they lunge and wrestle over ONE basketball — orange fire aura on Brunson vs silver-blue lightning aura on Wemby; (3) the gym doors EXPLODE open with a blue shockwave that blasts everyone off their feet BEFORE anyone enters; (4) five silhouettes stride in through the blinding doorway, revealed as LeBron James and the 76ers in royal blue; (5) close-up: LeBron raises a fist with FOUR championship rings and speaks the anime line. Background players (Anunoby #8, Towns #32 / Fox #4, Castle #5) may drift slightly — but Brunson and Wembanyama must be 100% consistent in every shot.
Production method — follow exactly:
Generate an anime style key + individual character sheets (seedream_v4_5, quality "basic") for Brunson, Wembanyama, and LeBron with correct real 2026-27 uniforms; checkpoint: show me the sheets for approval before continuing.
Generate the 5 keyframes (9:16) using the approved sheets as identity references (max 3 image refs per call; explicit height relations; "EXACTLY ONE basketball"; strict uniform rule: white jerseys only 11/8/32, black only 1/4/5). Fix small flaws via seedream self-edit: pass the frame back as reference + "reproduce EXACTLY, pixel-identical, ONE single change" (attach the character's sheet as a second ref for face fixes). Checkpoint: show me a labeled contact sheet of the 5 frames.
Write a short Japanese dialogue script (do NOT name any player in the lines): a taunt from Wemby, a defiant shout from Brunson, a ball-fight snarl, a terrified 「なんだ、この力は⁉」 at the shockwave, an awed 「まさか…あの男が来たのか…!」 whisper, and LeBron's finale 「お前はもう、死んでいる。」 Record with seed_audio using distinct male preset voices (get real voice IDs from list_voices — don't trust cached IDs); LeBron's line must be DEEP: pitch_rate -5, speech_rate -2.
Animate 5 clips on seedance_2_0 (1080p, std, 9:16, 4-5s each) with frame-chaining: each clip uses its keyframe as start_image AND the next shot's keyframe as end_image so the reel plays as one continuous take. For every clip where a character speaks on camera, upload the voice line as a silence-padded wav (media_upload + PUT + media_confirm) and pass it as audio_references with a prompt saying the character "speaks the line from the reference audio, mouth movements matching exactly" — never overlay dialogue in post over a non-speaking mouth. Use kinetic modern anime motion language: sakuga, crash zooms, whip pans, speed lines, impact frames, smear frames, screen shake, genre "action". Known traps: decline preset auto-recommendations with declined_preset_id; keep audio-ref chained clips at 4s (6s fails); if seedance fails silently twice on a pair, switch that clip to kling3_0 pro; if the content filter flags "menacing/fallen players" wording, soften and retry.
Assemble in the Higgsfield sandbox with ffmpeg: concat the 5 clips (normalize audio, add silent tracks where missing), keep clip-native audio at full volume (it carries the lip-synced dialogue), overlay the two off-camera voice lines at the right timestamps, and synthesize an instrumental battle-score bed procedurally with python in the sandbox (taiko build → boom on the door explosion → tension pulse → riser → dark held chord under the final line) — do NOT use any external music service. Upload the finished mp4 to my Higgsfield library and give me the link.
Throughout: run all downloads and QC yourself (frame extraction + visual checks), show me labeled preview images at every checkpoint, and tell me the credit cost before each generation phase.
Prompt for Heygen AI Avatar on Claude Code + Post on Blotato (IG)
Create a short AI avatar video with HeyGen, then post it to my socials with Blotato. Do this in stages and do not post anything until I approve the clip.
Stage 1 - Generate the clip
Use HeyGen's create_video_from_avatar with:
- avatarId: a761ce70b43447ab8383684d98afcf22 (stock avatar "Jibran", digital twin look)
- Use the avatar's default voice (omit voiceId)
- aspectRatio: 9:16, resolution: 1080p
- title: "Jibran explains Blotato"
- script, exactly as written: "If you're creating content but posting it everywhere is eating your time, look at Blotato. You create once, and it repurposes and publishes to all your social channels automatically. Less busywork, more creating."
Stage 2 - Review
Poll the video status until the render finishes, then give me the video link so I can watch it. Then STOP and ask me: "Approve this clip for posting?" Do not call any Blotato tool before I explicitly approve. If I ask for changes, regenerate and come back for approval again.
Stage 3 - Post with Blotato (only after my explicit approval)
1. List my connected Blotato accounts and tell me which ones you found.
2. Post the video to my Instagram and TikTok accounts with this caption: "This entire video was made and posted by AI with one prompt. HeyGen avatar plus Blotato for publishing.
3. Confirm when it's live and share the post link.
