Я Юзал Claude и Создал ИИ Модель для Монетизации

Product Reviewed (Implied)

The video is a walkthrough of using AI to create realistic “AI bloggers/avatars” (via Claude + Google/Flow + NanoBan, then video generation tools), and how to monetize them—especially with Yandex advertising for bloggers and YouTube monetization.

The “product” being pushed is essentially a workflow/tools stack, including:

• Claude (prompt/dialog optimization; can “generate prompts” and help with multilingual issues)
• Flow / Google video/image services, mentioning Veo 3.1 for video
• NanoBan / Nanobanana Pro / Nanobanana 2 (high-quality image generation + character consistency via reference-based replacement)
• Optional:
  • Photom (remove backgrounds)
  • Canva (collage / character reference sheet)
  • Video tools:
    • One (Creature Swap / Photo Animate; motion copy; “free credits” but with queues)
    • Hixwet (ad-style UGC video generation; text-to-speech; built-in translation)

---

Key Features / What It Can Do

1) Create a consistent AI blogger persona

• Start from a character reference image
• Generate multiple reference angles/face details to preserve identity across new images
• Focuses on solving the “consistent appearance” problem using multi-angle reference image prompts

2) Generate large volumes of high-quality photos

• Uses Flow + NanoBan
• Mentions free use of Nanobanana via Flow (“absolutely free of charge”)
• Supports 9:16 aspect ratio and multiple photos per run
• Mentions model options: NanoBan Pro vs Nanobanana 2 (“more or less the same”)

3) Customize pose, lighting, location, and clothes—while keeping the same character

• Create additional angles (side, back, eye close-ups) with reference images
• Add full-body/proportion prompts (height/body proportions, wings, moles/pimples, long arms/fingers, etc.)
• Iteratively improve results by adding/replacing references

4) Character reference “collage” pipeline for training/consistency

• Remove backgrounds with Photom
• Build a collage in Canva (canvas size given: 1920 × 1800)
• Use the collage as a “standard” to generate more images/videos of the same avatar

5) Video generation from a still avatar

• Uses Flow’s Veo 3.1:
  • 20 credits per short segment; Lite costs 10 credits
  • Mentions 50 credits free + 100 credits for registration
  • Example prompt: avatar talking naturally in Russian in a podcast studio, not looking at camera
• Uses One (Creature Swap / Photo Animate):
  • Motion copy from a reference dance video
  • Claimed 0 credits and “without limits,” but warns about long queues (often 7–10 minutes, sometimes 12–15 minutes)
  • In pro mode: preserves clothes/proportions/appearance while copying movement

6) Upscaling for better video quality

• Recommends upscaling images before video generation
• Example tool: image.upscaler
• Mentions 400% upscale to improve sharpness
• States differences may be subtle in stills but more noticeable in video

---

Monetization Options Described (Core Value Proposition)

1) Yandex Advertising Network for bloggers (YAN)

• Monetization requires at least 1,000 subscribers on Telegram/YouTube/other platforms
• Claims Yandex handles operational work:
  • Connect channel
  • Pick products/offers from hundreds of thousands of advertisers
  • Get paid for conversions (subscribe/purchase/add to cart), with reward credited later
• Workflow described:
  • Download product image
  • Generate an ad video with the avatar and matching scene (NanoBan + video generation)
  • Use Yandex neural network to write the post

2) Traditional “sell ads once” model

• Build audience first, then sell advertising placements
• Compared to YAN:
  • YAN can start earlier (claims “first video”)
  • YAN can earn continuously as conversions happen (per the description)

3) Trust / niche expert avatar model

• Example: “Rabbi Goldman” style account (influencer analogy)
• Requires earning trust, not just getting views
• Monetize via one link in profile selling a guide

4) YouTube monetization model

• Says they monetize via YouTube views, with example claims:
  • A channel network earned $7,000 in 3 months from YouTube views
  • Another claim: 8,000–10,000 views/day average
  • Some channels reaching ~$10,000/month
• Mentions a “team” and Razgon program for managing/growing channels or becoming a channel manager

---

Pros (As Presented)

• High character control and consistency (multi-angle references + collage standard)
• Can generate photos daily and multiple outputs to pick best
• Enables UGC-like ad creatives and promo videos quickly
• Monetization paths that don’t require building a huge audience first (especially via YAN)
• Motion:
  • “One” tool can animate using dance/movement references
  • Preserves appearance/clothes while copying movement
• Upscaling improves video results (important for avatar animation quality)

---

Cons / Limitations (As Presented)

• Starting setup is hard
  • Biggest challenge is consistent appearance, requiring reference-angle generation and collage building
• Free tier quality limits
  • Free download: up to 1K resolution
  • Limited higher: 2K resolution
  • Max quality requires subscription
• Language/voice issues
  • Hixwet described as understanding Russian poorly (producing mixed/garbled output); workaround is voicing in English first, then translating in the tool
• Queues for free animation
  • One can take 7–10 minutes, sometimes 12–15 minutes per generation
• YAN still depends on conversions
  • Not just views

---

Comparisons Made

• YAN vs classic sponsored ads
  • Classic: build audience first; paid per ad post
  • YAN: offers can be advertised immediately (claim: “in the first video”); reward can continue as conversions happen
• Veo 3.1 vs Lite model (within Flow)
  • Lite is cheaper (10 credits) and “often not much different” quality

---

Numerical Details / Ratings Mentioned

• YAN eligibility: 1,000 subscribers+
• Canva collage size: 1920 × 1800
• NanoBan downloads:
  • Free: 1K resolution
  • Limited: 2K resolution
  • Max quality: subscription
• Video generation (Flow / Veo 3.1):
  • 20 credits per segment
  • 50 credits free + 100 credits registration
  • Lite: 10 credits
• One (Creature Swap / Photo Animate):
  • Queue times: 7–10 min, sometimes 12–15 min
• YouTube claims:
  • Example: $7,000 in 3 months
  • View rate: 8,000–10,000 views/day
• Mentions another claim: “~10,000 Bachinskys per month” (currency unclear)

---

Unique Points Mentioned About the Workflow / Tooling (Consolidated)

• Create multiple “24/7 bloggers” that generate videos and advertise products
• Use Claude to craft optimized prompts for NanoBan (sometimes used to prepare prompts rather than generate images directly)
• Use Pinterest to find character references (choose one with clearly visible face)
• Combine multiple reference images to create a unique character
• Use Flow’s Nanobanana to generate “ultra-high-quality images daily” (as stated)
• Generate 9:16 format; generate multiple photos per model/run
• Achieve multi-angle consistency (side profile, back angle, eye close-up), emphasizing facial detection
• Add full-body/proportion prompts (arms, fingers, legs, skin features, wings, etc.)
• Iteratively fix artifacts by adding/replacing references (e.g., cross-eyed issues)
• Remove backgrounds with Photom for collage creation
• Use Canva to place side/front/eye images; keep eye close-up large/centered
• Upload the finished collage as reference for new scenes/outfits
• Example prompts include:
  • “walking in Miami, evening dress”
  • replacing a character in scene lighting/clothes
• YAN workflow: choose products/offers; get paid per actions/conversions; Yandex neural network assists writing
• Need videos for monetization reach (as stated)
• Upscale images before video generation (example 400%)
• Video generation options:
  • Flow / Veo 3.1 (credits, free credits)
  • One / Creature Swap or Photo Animate (free, but queues)
• Motion training:
  • Use a TikTok dance video + Telegram bot to save
  • Feed reference to One
• Ad video generation with Hixwet:
  • Choose CDEN 2.0 / Clink 3.0 (claimed best at the time)
  • Add product input; optionally attach voice
  • Use ElevenLabs for voice
  • Voice Russian in English, then translate because Russian may be garbled
  • UGC-style prompt example: phone-camera handheld perfume ad talking to camera
  • Ads video length up to 15 seconds (credits depend on duration/quality)
• Monetization models mentioned:
  • Ads-after-audience-growth
  • Trust-building niche expert avatar (single link selling a guide)
  • YouTube monetization using avatar channels managed by a team/program

---

Speaker Breakdown

• Only one main narrator appears in the subtitles.
• No clearly distinct multi-speaker viewpoints are separated.
• The video includes examples attributed to other creators (e.g., Aitana Lopes, Milo, Rabbi Goldman) plus the speaker’s own team results.

---

Concise Verdict / Recommendation

The video recommends a Claude + NanoBan/Flow + video/motion tools workflow for creating consistent AI blogger avatars and quickly producing promo content.

It strongly emphasizes that the hard part is setup—especially multi-angle character references and collage creation—and that quality/credits and queues can be limiting. However, it positions the monetization pathway—particularly YAN ads for bloggers with 1,000+ subscribers—as a major advantage over traditional sponsorships.

Overall: best for creators who want an automated avatar-driven content pipeline and plan to monetize via ads/conversions or YouTube. Less ideal if you want only a quick one-off avatar without investing in the consistency workflow.