Blog
Sume blog: agent-native video generation, Avatar 1.0 orchestration, and production APIs for builders and coding agents.
Sume error retry matrix: retry with the same key, or fix the request
Which Sume API errors to retry with the same Idempotency-Key, which to fix first, and which to leave alone. A table and a small Python function that decides.
Sume POST /v1/images (sume/auto) vs Image 1.0: which for new code
Sume's docs say Image 1.0 is retiring and is an alias for Image Router Auto. New code should use POST /v1/images. The field differences.
Synthesia Basic free 10 minutes vs 10 minutes of Sume avatar video
Synthesia Basic is $0 with 10 minutes and 9 avatars; the same 10 minutes on Sume standard is $110.40, so the free plan wins until you outgrow it.
TikTok ad bitrate window: 0.516 to 6.67 Mbps for 10 minutes
TikTok wants at least 516 kbps and at most 500 MB. For a 10-minute ad that is a 38.7 MB floor and a 6.67 Mbps average ceiling. The math and what Sume controls.
TikTok safe zones are files, Meta's are percentages: which to use
Meta gives Reels safe zones as 14% top, 35% bottom, 6% sides; TikTok gives downloadable files. How to build to both and place captions on Sume.
Sume TTS default is mp3: sentence slices need wav and emit_audio
Sume TTS defaults to mp3 at 44.1 kHz and 128 kbps. Per-sentence audio slices need wav (pcm_s16le) plus segmentation emit_audio, as in the Python request below.
YouTube swaps a 16:9 thumbnail on vertical videos: make 9:16 and 4:5
YouTube replaces 16:9 custom thumbnails on vertical videos with an auto 4:5 crop on home, explore and subscriptions. Two Sume images per Short cost $0.20 at 1K.
Which id to send Sume support: req_, job_, arun_ or agrun_
Error bodies carry req_ ids; jobs have job_; Format and Action runs arun_; Agent Completions agrun_. Which to share with support, and what never to paste.
YouTube AI label: player if photoreal, description if animated
YouTube shows the AI label in the player for photorealistic content and in the expanded description for animated content. Which Sume clips fall where.
YouTube AI disclosure exemptions: captions, color and audio repair
YouTube exempts captions, color adjustment, beauty filters, audio repair and own-voice dubs from AI disclosure. Which Sume jobs are only edits.
YouTube's Shorts AI tools self-label; a clip you upload does not
Clips from YouTube's own Shorts AI tools are disclosed automatically and run up to 10 seconds. A clip generated elsewhere and uploaded needs your own toggle.
Skippable in-stream: no max length, under 3 minutes recommended
Google Ads sets no maximum for skippable in-stream video but recommends under 3 minutes. A 5-minute master becomes a 179-second cut for $0.02 on Sume.
Amazon no-text lower-right rule vs Walmart contrast: where captions go
Amazon Sponsored Brands bans text in the lower right; Walmart wants 4.5:1 contrast. How to move Sume burned-in captions up with anchor ratios and check a still.
Walmart's three-flashes-a-second limit and Sume's 24-frame check
Walmart bans strobe and caps flashing at three times a second. Sume has no flash detector; video-frames returns up to 24 named stills to review one second.
Walmart music-only video: add a [MUSIC PLAYING] caption with Sume cues
Walmart's best practice for a music-only video is a [MUSIC PLAYING] caption, removed after about 10 seconds. Burn it with Sume caption cues for a flat $0.20.
Walmart silent video: the [NO SPEECH] caption and Kling audio off
A silent Walmart video should open with a [NO SPEECH] caption. Sume Kling 3 audio-off costs 14 cents a second; burning the caption adds a flat 20 cents.
Walmart's 4.5:1 text contrast for video: set Sume caption colors
Walmart wants 4.5:1 contrast for normal text and 3:1 above 18 px. Sume caption design.colors sets text and card colors; ratios for four pairs inside.
Walmart video .vtt captions: Sume burns text in, you build the file
Walmart calls closed captions optional but recommended and wants a .vtt name. Sume burns captions into pixels and returns no .vtt; here is the transcript route.
Photographer portfolio reel from stills with Omni Flash
Turn 10 portfolio photos into a 60-second reel with Gemini Omni Flash: six seconds each, $7.50 at 720p on Sume, with 360p drafts first at $2.25.
Gemini Omni rate limit: Google lists none, Sume lists them
Google's rate-limit page has no Omni or Veo row. It lists per-project RPM, TPM and RPD plus spend caps. Sume publishes per-key limits and concurrency by plan.
Roleplay team budget: Synthesia learner seats or a Sume clip library
Twelve-month cost of Synthesia Roleplay Training for 20 learners against a library of 24 scenario clips on Sume, and where each stops being the right tool.
Amazon video: key benefit in the first 10 seconds, checked in 5 stills
Amazon says the best shoppable videos show the key benefit in 5 to 10 seconds. Pull five stills at 1, 3, 5, 7 and 9 seconds with Sume video-frames to check.
Amazon video at 30, 60 or 90 seconds: Sume jobs per model and cost
Amazon says its best shoppable videos run 30 to 90 seconds. On Sume that is 3 to 9 jobs depending on the model, from $3.75 to $52.02 for 90 seconds.
Cheapest Sume image model for a 21:9 banner from a reference photo
Qwen Image at $0.025 is the cheapest row listing 21:9 with reference input. Flux 2 Pro is next at $0.0375. A short script finds it live.
Which Sume MCP tools to leave undeferred when Claude searches tools
Anthropic says to keep 3 to 5 frequently used tools non-deferred. Which Sume hosted MCP tools to pin and which to defer.
Claude tool search limits: 200-char regex, 500-char BM25, 5 results
Claude's tool search tool has fixed limits on pattern length, results and deferred tools. What they mean for a Sume hosted MCP tool list.
Deploy check: does your video model id and duration exist on Sume?
A Python preflight that reads GET /v1/videos/models and fails the deploy when VIDEO_MODEL is not a catalog id or the duration is outside supported_durations.
Env and secrets diff for removing Sora: SUME_API_KEY, one auth header
Which environment variables to delete, add and rotate when a service leaves the OpenAI Videos API for Sume, plus a startup check that fails on a missing secret.
Failed Sume video job: resubmit or stop? Read retryable, next_action
Sora said failed and little else. A failed Sume job carries an error with category, retryable, retry_after_seconds and next_action. A small decision function.
Sora content rules vs Sume generation_rejected: read the job events
A prompt Sora refused may behave differently on Sume. How a rejection shows up as generation_rejected, which events to read, and why you must not assume parity.
Instagram Reels with text over most of the frame: cap caption size
Instagram lists Reels mostly covered by text among those it may show less. Bound Sume captions with font_size_ratio, safe_width_ratio and max_words.
Reels ad safe zone in pixels: 358 top, 896 bottom, 86 sides
Meta's Reels ad page keeps about 14% top, 35% bottom and 6% per side clear on a 1440x2560 frame. Here is the pixel math and a Sume caption anchor.
LinkedIn Page video: 10 or 15 minutes, and is MOV allowed?
Two LinkedIn help pages give different Page video limits (10 vs 15 minutes, MOV in or out). Ship MP4 under 10 minutes and use Sume video trim to get there.
LinkedIn video uploads but never processes: ProRes fix
LinkedIn says a ProRes file may upload but not process. Probe the codec with Sume video inspect, then re-encode to H.264 MP4 with an exact video trim.
Museum exhibit teaser: Seedance 2.5, images and audio, no surcharge
Send artifact photos and a narration clip to Seedance 2.5 and a 10-second 720p teaser is $5.78 on Sume. Only reference videos trigger the higher bill.
Agent Completions model field is sume-agent only: no Haiku or GLM
You cannot choose Claude Haiku 5.5 or GLM 5.3 in the Agent Completions model field. Sume accepts sume-agent and returns 400 for anything else.
App onboarding clip from screenshots: Wan 3.0 takes 10 references
Send up to 10 app screens as references to Wan 3.0 and a 10-second 9:16 onboarding clip costs $0.63 at 480p or $1.25 at 720p on Sume.
Bookstore author event promo: Seedance 2.5 at 1:1, 480p, 8 s
A square 8-second author-event promo on Seedance 2.5 is $2.06 at 480p and $4.63 at 720p on Sume. Why 480p is enough for a feed, and what 1:1 renders at.
Boutique hotel room reel: three 8 s MiniMax H3 clips at 768p
Three 8-second 9:16 room clips on MiniMax H3 at 768p cost $0.60 each on Sume, $1.80 for the set. Plain H3 against H3 Max, and the reference limits.
Caption a MAI-Voice-2.1 narrated video: script_text keeps spelling
Burn captions on a clip voiced by MAI-Voice-2.1 or Flash: send the video URL and script_text, and Sume times the words. $0.20 for a clip up to 60 seconds.
Check ad video length for Pinterest, LinkedIn and Google in Python
A short Python check of a video-inspect probe against Pinterest, LinkedIn and Google Video action length limits read on 2026-10-08, before you upload an ad.
Claude MCP connector or Sume Agent Completions for a long video job?
A Messages API request with the MCP connector holds open while Claude works. Sume Agent Completions return 202 and a run to poll. Which one fits a long job.
ENABLE_TOOL_SEARCH=false in Claude Code loads every Sume tool up front
Setting ENABLE_TOOL_SEARCH=false or a custom ANTHROPIC_BASE_URL turns off Claude Code tool search. What that does to a Sume MCP session.
Claude Code tells Claude when an MCP server fails: Sume down vs auth
With tool search on, Claude Code reports failed MCP servers to the model. How to read that when Sume's server will not connect, 401 or 403.
Haiku 5.5 tool-use overhead: 286 tokens, 406 with tool_choice any
Claude Haiku 5.5 adds 286 system-prompt tokens when tools are present and 406 when tool_choice is any or tool. What that costs a Sume agent loop.
Claude per-message effort: drop to low while a Sume job runs
Claude's per-message effort beta keeps the prompt cache when you change effort mid-conversation. How to use it around Sume jobs, and its Haiku 5.5 limit.
Cloudflare Queue consumer 15 min wall time: poll a Sume job
A Queue consumer on Cloudflare can run 15 minutes of wall time, but a Sume job can wait longer. Re-enqueue with a delay and honor next_poll_after_seconds.
Convert MAI-Transcribe-2 word offsets to Sume video caption words
Turn word timings in milliseconds into the seconds-based words array Sume video-captions accepts. A short Python converter plus the 60 s and 1,200-word limits.