Sources, Updates & Corrections
Complete bibliography of primary technical documentation, segregated secondary commentary, date-stamped verifications, and editorial review processes.
Editorial Independence & Affiliation Notice
This site is not affiliated with, maintained by, or endorsed by Pika Labs, MiniMax AI, ByteDance, or Google.
We do not sell video subscriptions, host API keys, process payments, embed third-party tracking scripts, or participate in affiliate referral programs. All assessments are based strictly on publicly accessible technical documentation.
Primary Documentation Register
The technical parameters, node values, licensing rules, and pricing data across this guide derive from the following primary vendor documents, verified on October 30, 2026:
| Publisher / Author | Document Title | Direct Source Link | Access Date | Data Extracted |
|---|---|---|---|---|
| Pika Labs | Pika H3 Flex Official Product Page | pika.art/pika-h3-flex | October 30, 2026 | Per-second pricing tiers, launch promotion dates (Oct 29–Nov 5), 4K upscaling, and 5 WAV audio stems. |
| Pika Labs | Pika-H3-Flex Model Card (Hugging Face) | huggingface.co/Pika-Labs/Pika-H3-Flex | October 30, 2026 | 6-step distilled architecture, FL2VA and Ref2VA LoRA adapters, Apache 2.0 code license, MiniMax license inheritance. |
| Pika Labs | Pika-H3-Flex GitHub Repository | github.com/Pika-Labs/Pika-H3-Flex | October 30, 2026 | Python inference scripts, diffusers pipeline integration, and base repository download documentation. |
| Pika Labs | Pika Prompt Guidance Documentation | github.com/.../docs/prompt-guidance.md | October 30, 2026 | Official structural fields (subject_definitions:, overall_soundscape:, non_diegetic_music:), advisory camera rules, 0.2–0.3s temporal cut drift. |
| Pika Labs Developer | Pika Draft-to-Video Documentation | dev.pika.art/.../draft-to-video | October 30, 2026 | Two documented input paths: in-platform draft_job_id and external 480p/768p video media URL (distinguishing temporary PUT upload_url from persistent media url). |
| Pika Labs Developer | Pika Audio Stems & Pricing Documentation | dev.pika.art/.../audio-stems | October 30, 2026 | Two documented stem workflows (in-generation output_mode: audio-stems and standalone endpoint), sum-of-durations billing model ($0.001/sec), and 5-category WAV ZIP output. |
| Comfy-Org | Official ComfyUI Pika H3 Flex Tutorial | docs.comfy.org/tutorials/video/pika/pika-h3-flex | October 30, 2026 | ComfyUI version requirements (v0.30.0+ / v0.43.2+), model directories, and LCM sampler layout. |
| Comfy-Org | MiniMax-H3 Hugging Face Repository & Discussion #70 | huggingface.co/Comfy-Org/MiniMax-H3/discussions/70 | October 30, 2026 | Normalized ManualSigmas floats (1.0, 0.972973, 0.923077, 0.8, 0.571429, 0.333333, 0.0) and LoRA adapter weights. |
| MiniMax AI | MiniMax-H3 Model Card & Community License | huggingface.co/MiniMaxAI/MiniMax-H3/blob/main/LICENSE | October 30, 2026 | 33B foundation parameters, native 768p/2K resolution, territorial exclusions (EU, UK, KR, US), and prior written authorization requirement over $20M yearly revenue. |
| ByteDance Seed | Introducing Seedance 2.5: One-Take Creation & Flexible Referencing | seed.bytedance.com/.../introducing-seedance-2-5 | October 30, 2026 | 30-second one-take continuous video, up to 50 multimodal references, native synchronized audio, and region-level editing. |
| Gemini Omni Documentation & Video Capabilities | ai.google.dev/gemini-api/docs/omni | October 30, 2026 | 3–10s single generation, sequential multi-turn scene extensions up to 40 seconds, and conversational video editing via Interactions API. |
Secondary Commentary & Industry Analysis
The following third-party articles provide industry analysis and community perspectives. They are maintained separately from primary technical documentation:
| Publisher | Article Title | Link | Access Date | Classification |
|---|---|---|---|---|
| Oraclip | Pika H3 Flex: In-Depth Technical Breakdown | oraclip.com/blog/pika-h3-flex | October 30, 2026 | Independent secondary analysis and community overview. |
Editorial Corrections Policy
This project is committed to strict factual accuracy. Our correction route operates as follows:
Correction Workflow & Status
Current Review Stage: This site is currently a local review build. During this pre-deployment phase, factual checks, source validation, and code corrections are managed internally through the collaborative Codex/Gemini review workflow.
Post-Publishing Community Route: When the project owner designates a public hosting environment and public code repository, an official public issue tracker will be linked on this page. Community contributors will then be able to submit correction requests backed by primary documentation links.
Revision Changelog
| Date | Version | Revisions & Corrections |
|---|---|---|
| October 30, 2026 | v1.0.3 |
Upload schema & stem workflow clarification: Distinguished temporary PUT upload_url from persistent media url in upload flows; explained both documented stem workflows (in-generation output_mode: audio-stems vs. standalone endpoint); labeled stem table by category (Dialogue, Score, SFX, Foley, Ambience) without asserting unverified filenames; and reaffirmed untested status.
|
| October 30, 2026 | v1.0.2 |
API schema & licensing repair pass: Corrected draft-to-video guidance documenting both draft_job_id and external 480p/768p video URL upload paths; expanded audio-stems guide explaining standalone usage (input_url and request_id) and five-stem WAV ZIP output; formatted readable sum-of-durations billing calculation ($0.05 for five 10-second stems); faithfully summarized MiniMax Community License (no MAU metric, prior written authorization required over US $20 million yearly revenue); and qualified comparison metrics for Gemini Omni (internal schedule undisclosed) and Seedance 2.5 (vendor claim).
|
| October 30, 2026 | v1.0.1 |
Factual repair pass: Separated Pika CLI integer scheduler knots from ComfyUI normalized ManualSigmas (1.0 down to 0.0); corrected Comfy-Org Hugging Face link and version prerequisites; updated prompt templates to official field names; fixed HTML nesting; corrected MiniMax H3 native resolution (768p/2K) and Gemini Omni single-generation length (3–10s vs 40s extensions); and segregated secondary commentary.
|
| October 30, 2026 | v1.0.0 | Initial release with 6 connected pages, responsive layout, and documentation synthesis. |