Blog
Sume blog: agent-native video generation, Avatar 1.0 orchestration, and production APIs for builders and coding agents.
Grok Imagine on Sume lists 9:19.5 and 9:20 but not 4:5
Sume's Grok Imagine row lists tall phone ratios 9:19.5 and 9:20 and wide 20:9 and 19.5:9, but not 4:5, 5:4 or 21:9, and caps n at 1. Here is what to do.
A circuit breaker for paid video API calls: cap, poll, cancel
Guard paid video generation with three layers: a per-run spend cap, a loop breaker in your code, and cancel. Sume charges for work done before a cancel.
Billable vs debited on a Sume run: why the wallet moved more
On a Sume run, billable_amount_usd_micros is generation only; debited_usd_micros is the real wallet deduction and includes the LLM turn. Read both.
n8n Webhook node IP allowlist: Sume publishes no sender IPs
n8n can limit a Webhook trigger to listed IPs, but Sume's docs list no sender addresses. Verify the signature instead, with a Python check.
The 5 aspect ratios all 19 Sume image rows accept (a portable set)
All 19 Sume image rows list 1:1, 16:9, 9:16, 4:3 and 3:4. Use those five for a model bake-off and every row will accept the request without a 400.
ElevenLabs dubbing editor in maintenance mode: redo one line on Sume
ElevenLabs says its dubbing editor gets critical fixes only. If you need per-line regeneration, a Sume TTS job per sentence gives you one retry per line.
Write a Sume video URL back to a CMS record: what to store
Sume media URLs in a finished run are durable and public. Store them on your CMS record, and proxy or copy them if you need per-customer access control.
Grok Imagine Lite's 10 requests per second vs Sume plan concurrency
xAI lists a 10 requests per second limit and Batch API for Grok Imagine 1.5 Lite. On Sume, a clip batch is bounded by plan concurrency and queue capacity.
Sume API 401 on the dev host: keys only work on their own host
A Sume key works only on the host it was created for. A key from api.sume.com gets 401 on api.dev.sume.com and the reverse. Match key, host, and env var.
Poster headline in the image: Ideogram 4.5 or Nano Banana 2.1 on Sume?
Both vendors pitch readable text in images. On Sume, Ideogram 4.5 starts at $0.0375 and Nano Banana 2.1 at $0.075. Compare price, ratios and reference limits.
Nano Banana 2.1 or Pro for ad creative? 2.1 at 2K beats Pro at 1K
On Sume, Nano Banana 2.1 at 2K is $0.15, less than Pro at 1K ($0.1875). 2.1 also lists 1:4, 4:1, 1:8 and 8:1 banners that Pro lacks.
Can Imagen 4 use my product photo? No: pick an edit-capable row
Imagen 4 Fast and Ultra on Sume are text-to-image only and reject references. Google also says Imagen is shut down in its API. Six rows take a product photo.
Pop-up shop teasers: one still, three 5-second videos, dated cues
Plan a three-day pop-up shop teaser series on Sume: one Nano Banana 2.1 still, three Wan 3.0 jobs, three caption cue jobs. Chain and limits, read 2026-10-10.
Meta image expansion or a native 9:16 Sume image for Reels ads?
Meta can expand a creative to 9:16 with generative AI for Reels ads. A native 9:16 image from Sume keeps the composition you chose. When each fits.
Launch teaser: one Seedance 2.5 hero shot plus two Wan fillers
Spend on one 5-second Seedance 2.5 hero shot at 720p ($2.889) and fill the rest with cheap Wan 3.0 480p. Timeline chain and a $3.69 total, read 2026-10-10.
Which Format version ran my API call? Check the receipt
Every Sume Format run receipt carries format.version. Read it after you edit a package, so a bulk batch or schedule is never judged against the wrong version.
Guardrails for an AI agent that calls paid media APIs
Five guardrails for an agent that spends money: required cap, read before paying, idempotency keys, no secrets in logs, and a cancel path. Drawn from Sume docs.
Scribe v2 audio-event tags and 32 speakers vs Sume STT's fixed flags
ElevenLabs Scribe v2 tags audio events and separates up to 32 speakers. Sume STT fixes diarize and tag_audio_events server-side, so those toggles do not exist.
Luma Ray 2 5-second keyframes: Sume rows with start and end frames
Luma's API makes Ray 2 clips up to 5 s from start and end keyframes. Sume has no Luma row; these Sume rows accept both frames and cover a 5-second clip.
German voice-over on Sume: TTS language field plus cue captions
Localize a 300-character promo into German with Sume TTS and cue captions for $0.31. What the language field covers, read 2026-10-10.
Hedra batch_size 1-8 variations vs Sume's one job per request
Hedra's API returns 1 to 8 variations per request; Sume avatar videos are one job each. How to get variations on Sume with previews and keys (read 2026-10-10).
Move a Vidu Q4 job to Sume: six fields to check before you submit
Vidu Q4 Preview's image-to-video fields (duration, resolution, start image, audio) mapped to Sume's /v1/videos names, with a Python filter for rows that fit.
D-ID audio talks: 15 MB, 5-10 minutes vs Sume's 4-60 s script
D-ID takes up to 40,000 characters of text or 15 MB of audio for a talk. Sume takes an English script of 4 to 60 seconds. Which input fits (read 2026-10-10).
Gemini TTS has 30 prebuilt voices; Sume TTS takes avatar voice ids
Google's Gemini TTS page lists 30 prebuilt voices. Sume TTS takes a voice UUID or voi_ id, or an avatar whose voice.status is ready; cloning stays app-only.
Fit an AI voiceover to a target length with Sume TTS speed 0.6 to 1.5
To hit a target duration, generate once, measure the audio, then set generation_config.speed to measured over target, clamped to Sume's 0.6 to 1.5 range.
Sume TTS 400 tts_language_script_mismatch: Korean encoding check
Sume TTS rejects language ko when the transcript has no Hangul syllable: 400 tts_language_script_mismatch, no charge. Usually the text was decoded wrongly.
Voxtral Mini 4B Realtime Arabic: open weights vs a Sume STT job
Mistral's Arabic streaming model arrived Oct 8 as Apache 2.0 weights, not a service. For Arabic transcripts without hosting, Sume STT takes language_code ar.
ElevenLabs 8 kHz mu-law and A-law vs Sume TTS pcm_mulaw, pcm_alaw
For phone-line audio, ElevenLabs lists 8 kHz mu-law and A-law. Sume TTS 1.0 offers pcm_mulaw and pcm_alaw in a raw container at 8000 Hz.
Scribe accepts MP4 and MOV; Sume STT needs an audio URL: detach first
ElevenLabs Scribe takes video files such as MP4 and MOV directly. Sume STT wants a public HTTPS audio_url, so a video goes through audio detach first at $0.01.
OpenAI TTS has Opus, AAC and FLAC; Sume TTS has MP3, WAV and raw
OpenAI's speech API lists six output formats. Sume TTS 1.0 returns MP3, WAV or raw PCM, so Opus, AAC and FLAC need a conversion step after generation.
Sume STT words[] cap: what words_truncated and words_total mean
Sume STT returns word timings capped at 20,000 entries and sets words_truncated and words_total when it hits the cap. A 600-second job stays far below it.
MAI-Voice-2.1 pricing vs Sume TTS: what Sume lists, what it does not
MAI-Voice-2.1 costs $22 per 1M characters per Microsoft AI. Sume does not list MAI; its TTS runs Sonic voices at $47.50 per 1M characters.
Silent reference clip: stt_skipped_silent, no-audio and caption errors
A silent clip does not fail a Sume reference ingest, and STT settles to zero. Video inspect and captions do raise errors on silence.
Veo 3.1 request on Sume: 720p, 4, 6 or 8 seconds, in Python
A working POST /v1/videos call for veo-3.1-lite with 6 seconds, 9:16 and no audio, the rates for the three Veo rows, and what Sume rejects.
Veo 3.1 Lite (Preview) on Sume: what the preview label means
Sume's catalog names the row 'Veo 3.1 Lite (Preview)'. Its limits, the audio rates, and how it differs from the Gemini API's Lite lines for 720p and 1080p.
Veo 3.1 on Sume: the audio-on rate equals Gemini API's 720p list
Sume's Veo audio-on list rates are $0.40, $0.10 and $0.05 per second, the same as Gemini API's 720p lines; then Sume bills times 1.25. What you pay and why.
ElevenLabs Music is cleared for nearly all commercial uses: verify
ElevenLabs says Eleven Music is cleared for nearly all commercial uses, with plan differences in separate terms. Check them against your use before publishing.
ElevenLabs Music 3 s to 5 min vs Sume Music: how to set length
ElevenLabs Music takes 3 seconds to 5 minutes. Sume Music has no duration field and rejects one: set length in the prompt or trim the result afterwards.
Suno Italy probe: four terms clauses to check before ad music
Italy's AGCM opened a probe into Suno's terms on Oct 6, 2026. The four flagged clauses, and what to check in any AI music tool before putting a track in an ad.
Media API length limits side by side: trim, filter, frames, compose
One table of the source and output length limits for Sume's media tools: 300 s, 900 s, 1800 s and the 60 s caption price band. Read 2026-10-10.
Porting chat-completions code to Sume Agent Completions
Agent Completions takes system and user messages, rejects assistant turns, returns a 202 receipt, and does not stream. Here is what to change when porting.
How do I ask Sume for 2K? Only 5 of 19 image rows list resolution
Only Nano Banana 2.1, Nano Banana Pro, Imagen 4 Ultra, Ideogram 4.5 and Soul list a resolution field. Send it to FLUX, Seedream or GPT and you get a 400.
GitHub Actions matrix cap is 256 jobs: use Sume bulk runs instead
A GitHub Actions matrix can create 256 jobs, and a Sume bulk run takes 100 items. How to split a batch between them so concurrency and queue limits hold.
Sume webhook URL localhost returns 400: test with a tunnel
Sume refuses localhost, private ranges, and plain HTTP webhook URLs with a 400 at create. Use a public HTTPS tunnel and verify the signature.
A sensor for a Sume Format run: poll, back off, time out
Poll a Sume Format run with a doubling backoff up to 60 seconds and a hard deadline. Here is the loop, the terminal states, and a 45 minute limit.
Which language does Lyria 3.5 sing in? Spanish prompt on Sume Music
Google says Lyria generates music in the language of your prompt. On Sume, write the brief in the language you want sung; no language field exists on Music.
Prove a Sume retry makes one job: drop the response on purpose
A 28-line fake server creates the job, then hangs up before replying. A stable Idempotency-Key leaves one job; a fresh key per retry leaves two. Runs anywhere.
How many Format runs per minute can my Sume plan start?
Sume limits writes per minute by plan, from 120 on Free to 1200 on Scale, with reads at 40 times that. Read the rate-limit headers and back off on 429.